diff options
Diffstat (limited to 'arch/ia64/sn/kernel')
26 files changed, 1983 insertions, 7084 deletions
diff --git a/arch/ia64/sn/kernel/Makefile b/arch/ia64/sn/kernel/Makefile index ab9c48c8801..d27df1d45da 100644 --- a/arch/ia64/sn/kernel/Makefile +++ b/arch/ia64/sn/kernel/Makefile @@ -4,18 +4,15 @@ # License. See the file "COPYING" in the main directory of this archive # for more details. # -# Copyright (C) 1999,2001-2005 Silicon Graphics, Inc. All Rights Reserved. +# Copyright (C) 1999,2001-2006,2008 Silicon Graphics, Inc. All Rights Reserved. # -CPPFLAGS += -I$(srctree)/arch/ia64/sn/include +ccflags-y := -Iarch/ia64/sn/include obj-y += setup.o bte.o bte_error.o irq.o mca.o idle.o \ - huberror.o io_init.o iomv.o klconflib.o pio_phys.o \ + huberror.o io_acpi_init.o io_common.o \ + io_init.o iomv.o klconflib.o pio_phys.o \ sn2/ obj-$(CONFIG_IA64_GENERIC) += machvec.o obj-$(CONFIG_SGI_TIOCX) += tiocx.o -obj-$(CONFIG_IA64_SGI_SN_XP) += xp.o -xp-y := xp_main.o xp_nofault.o -obj-$(CONFIG_IA64_SGI_SN_XP) += xpc.o -xpc-y := xpc_main.o xpc_channel.o xpc_partition.o -obj-$(CONFIG_IA64_SGI_SN_XP) += xpnet.o +obj-$(CONFIG_PCI_MSI) += msi_sn.o diff --git a/arch/ia64/sn/kernel/bte.c b/arch/ia64/sn/kernel/bte.c index 7f73ad4408a..cad775a1a15 100644 --- a/arch/ia64/sn/kernel/bte.c +++ b/arch/ia64/sn/kernel/bte.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (c) 2000-2006 Silicon Graphics, Inc. All Rights Reserved. + * Copyright (c) 2000-2007 Silicon Graphics, Inc. All Rights Reserved. */ #include <linux/module.h> @@ -19,6 +19,7 @@ #include <linux/bootmem.h> #include <linux/string.h> #include <linux/sched.h> +#include <linux/slab.h> #include <asm/sn/bte.h> @@ -63,7 +64,7 @@ static inline void bte_start_transfer(struct bteinfo_s *bte, u64 len, u64 mode) * Use the block transfer engine to move kernel memory from src to dest * using the assigned mode. * - * Paramaters: + * Parameters: * src - physical address of the transfer source. * dest - physical address of the transfer destination. * len - number of bytes to transfer from source to dest. @@ -97,9 +98,10 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification) return BTE_SUCCESS; } - BUG_ON((len & L1_CACHE_MASK) || - (src & L1_CACHE_MASK) || (dest & L1_CACHE_MASK)); - BUG_ON(!(len < ((BTE_LEN_MASK + 1) << L1_CACHE_SHIFT))); + BUG_ON(len & L1_CACHE_MASK); + BUG_ON(src & L1_CACHE_MASK); + BUG_ON(dest & L1_CACHE_MASK); + BUG_ON(len > BTE_MAX_XFER); /* * Start with interface corresponding to cpu number @@ -227,7 +229,7 @@ retry_bteop: BTE_LNSTAT_LOAD(bte), *bte->most_rcnt_na)); if (transfer_stat & IBLS_ERROR) { - bte_status = transfer_stat & ~IBLS_ERROR; + bte_status = BTE_GET_ERROR_STATUS(transfer_stat); } else { bte_status = BTE_SUCCESS; } @@ -247,7 +249,7 @@ EXPORT_SYMBOL(bte_copy); * use the block transfer engine to move kernel * memory from src to dest using the assigned mode. * - * Paramaters: + * Parameters: * src - physical address of the transfer source. * dest - physical address of the transfer destination. * len - number of bytes to transfer from source to dest. @@ -255,7 +257,7 @@ EXPORT_SYMBOL(bte_copy); * for IBCT0/1 in the SGI documentation. * * NOTE: If the source, dest, and len are all cache line aligned, - * then it would be _FAR_ preferrable to use bte_copy instead. + * then it would be _FAR_ preferable to use bte_copy instead. */ bte_result_t bte_unaligned_copy(u64 src, u64 dest, u64 len, u64 mode) { @@ -300,7 +302,7 @@ bte_result_t bte_unaligned_copy(u64 src, u64 dest, u64 len, u64 mode) * a standard bte copy. * * One nasty exception to the above rule is when the - * source and destination are not symetrically + * source and destination are not symmetrically * mis-aligned. If the source offset from the first * cache line is different from the destination offset, * we make the first section be the entire transfer @@ -337,7 +339,7 @@ bte_result_t bte_unaligned_copy(u64 src, u64 dest, u64 len, u64 mode) if (footBcopyDest == (headBcopyDest + headBcopyLen)) { /* - * We have two contigous bcopy + * We have two contiguous bcopy * blocks. Merge them. */ headBcopyLen += footBcopyLen; @@ -375,20 +377,19 @@ bte_result_t bte_unaligned_copy(u64 src, u64 dest, u64 len, u64 mode) } else { /* - * The transfer is not symetric, we will + * The transfer is not symmetric, we will * allocate a buffer large enough for all the * data, bte_copy into that buffer and then * bcopy to the destination. */ - /* Add the leader from source */ - headBteLen = len + (src & L1_CACHE_MASK); - /* Add the trailing bytes from footer. */ - headBteLen += L1_CACHE_BYTES - (headBteLen & L1_CACHE_MASK); - headBteSource = src & ~L1_CACHE_MASK; headBcopySrcOffset = src & L1_CACHE_MASK; headBcopyDest = dest; headBcopyLen = len; + + headBteSource = src - headBcopySrcOffset; + /* Add the leading and trailing bytes from source */ + headBteLen = L1_CACHE_ALIGN(len + headBcopySrcOffset); } if (headBcopyLen > 0) { diff --git a/arch/ia64/sn/kernel/bte_error.c b/arch/ia64/sn/kernel/bte_error.c index f1ec1370b3e..4cb09f3f1ef 100644 --- a/arch/ia64/sn/kernel/bte_error.c +++ b/arch/ia64/sn/kernel/bte_error.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (c) 2000-2005 Silicon Graphics, Inc. All Rights Reserved. + * Copyright (c) 2000-2007 Silicon Graphics, Inc. All Rights Reserved. */ #include <linux/types.h> @@ -78,7 +78,7 @@ int shub1_bte_error_handler(unsigned long _nodepda) * There are errors which still need to be cleaned up by * hubiio_crb_error_handler */ - mod_timer(recovery_timer, HZ * 5); + mod_timer(recovery_timer, jiffies + (HZ * 5)); BTE_PRINTK(("eh:%p:%d Marked Giving up\n", err_nodepda, smp_processor_id())); return 1; @@ -95,7 +95,7 @@ int shub1_bte_error_handler(unsigned long _nodepda) icrbd.ii_icrb0_d_regval = REMOTE_HUB_L(nasid, IIO_ICRB_D(i)); if (icrbd.d_bteop) { - mod_timer(recovery_timer, HZ * 5); + mod_timer(recovery_timer, jiffies + (HZ * 5)); BTE_PRINTK(("eh:%p:%d Valid %d, Giving up\n", err_nodepda, smp_processor_id(), i)); @@ -105,7 +105,7 @@ int shub1_bte_error_handler(unsigned long _nodepda) } BTE_PRINTK(("eh:%p:%d Cleaning up\n", err_nodepda, smp_processor_id())); - /* Reenable both bte interfaces */ + /* Re-enable both bte interfaces */ imem.ii_imem_regval = REMOTE_HUB_L(nasid, IIO_IMEM); imem.ii_imem_fld_s.i_b0_esd = imem.ii_imem_fld_s.i_b1_esd = 1; REMOTE_HUB_S(nasid, IIO_IMEM, imem.ii_imem_regval); @@ -148,9 +148,13 @@ int shub2_bte_error_handler(unsigned long _nodepda) for (i = 0; i < BTES_PER_NODE; i++) { bte = &err_nodepda->bte_if[i]; status = BTE_LNSTAT_LOAD(bte); - if ((status & IBLS_ERROR) || !(status & IBLS_BUSY)) + if (status & IBLS_ERROR) { + bte->bh_error = BTE_SHUB2_ERROR(status); continue; - mod_timer(recovery_timer, HZ * 5); + } + if (!(status & IBLS_BUSY)) + continue; + mod_timer(recovery_timer, jiffies + (HZ * 5)); BTE_PRINTK(("eh:%p:%d Marked Giving up\n", err_nodepda, smp_processor_id())); return 1; @@ -243,7 +247,7 @@ bte_crb_error_handler(cnodeid_t cnode, int btenum, /* * The caller has already figured out the error type, we save that - * in the bte handle structure for the thread excercising the + * in the bte handle structure for the thread exercising the * interface to consume. */ bte->bh_error = ioe->ie_errortype + BTEFAIL_OFFSET; diff --git a/arch/ia64/sn/kernel/huberror.c b/arch/ia64/sn/kernel/huberror.c index 96fb81e6321..f925dec2da9 100644 --- a/arch/ia64/sn/kernel/huberror.c +++ b/arch/ia64/sn/kernel/huberror.c @@ -3,12 +3,11 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 1992 - 1997, 2000,2002-2005 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 1992 - 1997, 2000,2002-2007 Silicon Graphics, Inc. All rights reserved. */ #include <linux/types.h> #include <linux/interrupt.h> -#include <linux/pci.h> #include <asm/delay.h> #include <asm/sn/sn_sal.h> #include "ioerror.h" @@ -22,7 +21,7 @@ void hubiio_crb_error_handler(struct hubdev_info *hubdev_info); extern void bte_crb_error_handler(cnodeid_t, int, int, ioerror_t *, int); -static irqreturn_t hub_eint_handler(int irq, void *arg, struct pt_regs *ep) +static irqreturn_t hub_eint_handler(int irq, void *arg) { struct hubdev_info *hubdev_info; struct ia64_sal_retval ret_stuff; @@ -38,12 +37,20 @@ static irqreturn_t hub_eint_handler(int irq, void *arg, struct pt_regs *ep) (u64) nasid, 0, 0, 0, 0, 0, 0); if ((int)ret_stuff.v0) - panic("hubii_eint_handler(): Fatal TIO Error"); + panic("%s: Fatal %s Error", __func__, + ((nasid & 1) ? "TIO" : "HUBII")); if (!(nasid & 1)) /* Not a TIO, handle CRB errors */ (void)hubiio_crb_error_handler(hubdev_info); - } else - bte_error_handler((unsigned long)NODEPDA(nasid_to_cnodeid(nasid))); + } else + if (nasid & 1) { /* TIO errors */ + SAL_CALL_NOLOCK(ret_stuff, SN_SAL_HUB_ERROR_INTERRUPT, + (u64) nasid, 0, 0, 0, 0, 0, 0); + + if ((int)ret_stuff.v0) + panic("%s: Fatal TIO Error", __func__); + } else + bte_error_handler((unsigned long)NODEPDA(nasid_to_cnodeid(nasid))); return IRQ_HANDLED; } @@ -178,11 +185,15 @@ void hubiio_crb_error_handler(struct hubdev_info *hubdev_info) */ void hub_error_init(struct hubdev_info *hubdev_info) { - if (request_irq(SGI_II_ERROR, (void *)hub_eint_handler, IRQF_SHARED, - "SN_hub_error", (void *)hubdev_info)) - printk("hub_error_init: Failed to request_irq for 0x%p\n", + + if (request_irq(SGI_II_ERROR, hub_eint_handler, IRQF_SHARED, + "SN_hub_error", hubdev_info)) { + printk(KERN_ERR "hub_error_init: Failed to request_irq for 0x%p\n", hubdev_info); - return; + return; + } + irq_set_handler(SGI_II_ERROR, handle_level_irq); + sn_set_err_irq_affinity(SGI_II_ERROR); } @@ -195,11 +206,15 @@ void hub_error_init(struct hubdev_info *hubdev_info) */ void ice_error_init(struct hubdev_info *hubdev_info) { + if (request_irq (SGI_TIO_ERROR, (void *)hub_eint_handler, IRQF_SHARED, "SN_TIO_error", - (void *)hubdev_info)) + (void *)hubdev_info)) { printk("ice_error_init: request_irq() error hubdev_info 0x%p\n", hubdev_info); - return; + return; + } + irq_set_handler(SGI_TIO_ERROR, handle_level_irq); + sn_set_err_irq_affinity(SGI_TIO_ERROR); } diff --git a/arch/ia64/sn/kernel/io_acpi_init.c b/arch/ia64/sn/kernel/io_acpi_init.c new file mode 100644 index 00000000000..0640739cc20 --- /dev/null +++ b/arch/ia64/sn/kernel/io_acpi_init.c @@ -0,0 +1,510 @@ +/* + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2006 Silicon Graphics, Inc. All rights reserved. + */ + +#include <asm/sn/types.h> +#include <asm/sn/addrs.h> +#include <asm/sn/pcidev.h> +#include <asm/sn/pcibus_provider_defs.h> +#include <asm/sn/sn_sal.h> +#include "xtalk/hubdev.h" +#include <linux/acpi.h> +#include <linux/slab.h> +#include <linux/export.h> + + +/* + * The code in this file will only be executed when running with + * a PROM that has ACPI IO support. (i.e., SN_ACPI_BASE_SUPPORT() == 1) + */ + + +/* + * This value must match the UUID the PROM uses + * (io/acpi/defblk.c) when building a vendor descriptor. + */ +struct acpi_vendor_uuid sn_uuid = { + .subtype = 0, + .data = { 0x2c, 0xc6, 0xa6, 0xfe, 0x9c, 0x44, 0xda, 0x11, + 0xa2, 0x7c, 0x08, 0x00, 0x69, 0x13, 0xea, 0x51 }, +}; + +struct sn_pcidev_match { + u8 bus; + unsigned int devfn; + acpi_handle handle; +}; + +/* + * Perform the early IO init in PROM. + */ +static long +sal_ioif_init(u64 *result) +{ + struct ia64_sal_retval isrv = {0,0,0,0}; + + SAL_CALL_NOLOCK(isrv, + SN_SAL_IOIF_INIT, 0, 0, 0, 0, 0, 0, 0); + *result = isrv.v0; + return isrv.status; +} + +/* + * sn_acpi_hubdev_init() - This function is called by acpi_ns_get_device_callback() + * for all SGIHUB and SGITIO acpi devices defined in the + * DSDT. It obtains the hubdev_info pointer from the + * ACPI vendor resource, which the PROM setup, and sets up the + * hubdev_info in the pda. + */ + +static acpi_status __init +sn_acpi_hubdev_init(acpi_handle handle, u32 depth, void *context, void **ret) +{ + struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + struct acpi_buffer name_buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + u64 addr; + struct hubdev_info *hubdev; + struct hubdev_info *hubdev_ptr; + int i; + u64 nasid; + struct acpi_resource *resource; + acpi_status status; + struct acpi_resource_vendor_typed *vendor; + extern void sn_common_hubdev_init(struct hubdev_info *); + + status = acpi_get_vendor_resource(handle, METHOD_NAME__CRS, + &sn_uuid, &buffer); + if (ACPI_FAILURE(status)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR + "sn_acpi_hubdev_init: acpi_get_vendor_resource() " + "(0x%x) failed for: %s\n", status, + (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return AE_OK; /* Continue walking namespace */ + } + + resource = buffer.pointer; + vendor = &resource->data.vendor_typed; + if ((vendor->byte_length - sizeof(struct acpi_vendor_uuid)) != + sizeof(struct hubdev_info *)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR + "sn_acpi_hubdev_init: Invalid vendor data length: " + "%d for: %s\n", + vendor->byte_length, (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + goto exit; + } + + memcpy(&addr, vendor->byte_data, sizeof(struct hubdev_info *)); + hubdev_ptr = __va((struct hubdev_info *) addr); + + nasid = hubdev_ptr->hdi_nasid; + i = nasid_to_cnodeid(nasid); + hubdev = (struct hubdev_info *)(NODEPDA(i)->pdinfo); + *hubdev = *hubdev_ptr; + sn_common_hubdev_init(hubdev); + +exit: + kfree(buffer.pointer); + return AE_OK; /* Continue walking namespace */ +} + +/* + * sn_get_bussoft_ptr() - The pcibus_bussoft pointer is found in + * the ACPI Vendor resource for this bus. + */ +static struct pcibus_bussoft * +sn_get_bussoft_ptr(struct pci_bus *bus) +{ + u64 addr; + struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + struct acpi_buffer name_buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + acpi_handle handle; + struct pcibus_bussoft *prom_bussoft_ptr; + struct acpi_resource *resource; + acpi_status status; + struct acpi_resource_vendor_typed *vendor; + + + handle = acpi_device_handle(PCI_CONTROLLER(bus)->companion); + status = acpi_get_vendor_resource(handle, METHOD_NAME__CRS, + &sn_uuid, &buffer); + if (ACPI_FAILURE(status)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR "%s: " + "acpi_get_vendor_resource() failed (0x%x) for: %s\n", + __func__, status, (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return NULL; + } + resource = buffer.pointer; + vendor = &resource->data.vendor_typed; + + if ((vendor->byte_length - sizeof(struct acpi_vendor_uuid)) != + sizeof(struct pcibus_bussoft *)) { + printk(KERN_ERR + "%s: Invalid vendor data length %d\n", + __func__, vendor->byte_length); + kfree(buffer.pointer); + return NULL; + } + memcpy(&addr, vendor->byte_data, sizeof(struct pcibus_bussoft *)); + prom_bussoft_ptr = __va((struct pcibus_bussoft *) addr); + kfree(buffer.pointer); + + return prom_bussoft_ptr; +} + +/* + * sn_extract_device_info - Extract the pcidev_info and the sn_irq_info + * pointers from the vendor resource using the + * provided acpi handle, and copy the structures + * into the argument buffers. + */ +static int +sn_extract_device_info(acpi_handle handle, struct pcidev_info **pcidev_info, + struct sn_irq_info **sn_irq_info) +{ + u64 addr; + struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + struct acpi_buffer name_buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + struct sn_irq_info *irq_info, *irq_info_prom; + struct pcidev_info *pcidev_ptr, *pcidev_prom_ptr; + struct acpi_resource *resource; + int ret = 0; + acpi_status status; + struct acpi_resource_vendor_typed *vendor; + + /* + * The pointer to this device's pcidev_info structure in + * the PROM, is in the vendor resource. + */ + status = acpi_get_vendor_resource(handle, METHOD_NAME__CRS, + &sn_uuid, &buffer); + if (ACPI_FAILURE(status)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR + "%s: acpi_get_vendor_resource() failed (0x%x) for: %s\n", + __func__, status, (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return 1; + } + + resource = buffer.pointer; + vendor = &resource->data.vendor_typed; + if ((vendor->byte_length - sizeof(struct acpi_vendor_uuid)) != + sizeof(struct pci_devdev_info *)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR + "%s: Invalid vendor data length: %d for: %s\n", + __func__, vendor->byte_length, + (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + ret = 1; + goto exit; + } + + pcidev_ptr = kzalloc(sizeof(struct pcidev_info), GFP_KERNEL); + if (!pcidev_ptr) + panic("%s: Unable to alloc memory for pcidev_info", __func__); + + memcpy(&addr, vendor->byte_data, sizeof(struct pcidev_info *)); + pcidev_prom_ptr = __va(addr); + memcpy(pcidev_ptr, pcidev_prom_ptr, sizeof(struct pcidev_info)); + + /* Get the IRQ info */ + irq_info = kzalloc(sizeof(struct sn_irq_info), GFP_KERNEL); + if (!irq_info) + panic("%s: Unable to alloc memory for sn_irq_info", __func__); + + if (pcidev_ptr->pdi_sn_irq_info) { + irq_info_prom = __va(pcidev_ptr->pdi_sn_irq_info); + memcpy(irq_info, irq_info_prom, sizeof(struct sn_irq_info)); + } + + *pcidev_info = pcidev_ptr; + *sn_irq_info = irq_info; + +exit: + kfree(buffer.pointer); + return ret; +} + +static unsigned int +get_host_devfn(acpi_handle device_handle, acpi_handle rootbus_handle) +{ + unsigned long long adr; + acpi_handle child; + unsigned int devfn; + int function; + acpi_handle parent; + int slot; + acpi_status status; + struct acpi_buffer name_buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + + acpi_get_name(device_handle, ACPI_FULL_PATHNAME, &name_buffer); + + /* + * Do an upward search to find the root bus device, and + * obtain the host devfn from the previous child device. + */ + child = device_handle; + while (child) { + status = acpi_get_parent(child, &parent); + if (ACPI_FAILURE(status)) { + printk(KERN_ERR "%s: acpi_get_parent() failed " + "(0x%x) for: %s\n", __func__, status, + (char *)name_buffer.pointer); + panic("%s: Unable to find host devfn\n", __func__); + } + if (parent == rootbus_handle) + break; + child = parent; + } + if (!child) { + printk(KERN_ERR "%s: Unable to find root bus for: %s\n", + __func__, (char *)name_buffer.pointer); + BUG(); + } + + status = acpi_evaluate_integer(child, METHOD_NAME__ADR, NULL, &adr); + if (ACPI_FAILURE(status)) { + printk(KERN_ERR "%s: Unable to get _ADR (0x%x) for: %s\n", + __func__, status, (char *)name_buffer.pointer); + panic("%s: Unable to find host devfn\n", __func__); + } + + kfree(name_buffer.pointer); + + slot = (adr >> 16) & 0xffff; + function = adr & 0xffff; + devfn = PCI_DEVFN(slot, function); + return devfn; +} + +/* + * find_matching_device - Callback routine to find the ACPI device + * that matches up with our pci_dev device. + * Matching is done on bus number and devfn. + * To find the bus number for a particular + * ACPI device, we must look at the _BBN method + * of its parent. + */ +static acpi_status +find_matching_device(acpi_handle handle, u32 lvl, void *context, void **rv) +{ + unsigned long long bbn = -1; + unsigned long long adr; + acpi_handle parent = NULL; + acpi_status status; + unsigned int devfn; + int function; + int slot; + struct sn_pcidev_match *info = context; + struct acpi_buffer name_buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + + status = acpi_evaluate_integer(handle, METHOD_NAME__ADR, NULL, + &adr); + if (ACPI_SUCCESS(status)) { + status = acpi_get_parent(handle, &parent); + if (ACPI_FAILURE(status)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR + "%s: acpi_get_parent() failed (0x%x) for: %s\n", + __func__, status, (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return AE_OK; + } + status = acpi_evaluate_integer(parent, METHOD_NAME__BBN, + NULL, &bbn); + if (ACPI_FAILURE(status)) { + acpi_get_name(handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR + "%s: Failed to find _BBN in parent of: %s\n", + __func__, (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return AE_OK; + } + + slot = (adr >> 16) & 0xffff; + function = adr & 0xffff; + devfn = PCI_DEVFN(slot, function); + if ((info->devfn == devfn) && (info->bus == bbn)) { + /* We have a match! */ + info->handle = handle; + return 1; + } + } + return AE_OK; +} + +/* + * sn_acpi_get_pcidev_info - Search ACPI namespace for the acpi + * device matching the specified pci_dev, + * and return the pcidev info and irq info. + */ +int +sn_acpi_get_pcidev_info(struct pci_dev *dev, struct pcidev_info **pcidev_info, + struct sn_irq_info **sn_irq_info) +{ + unsigned int host_devfn; + struct sn_pcidev_match pcidev_match; + acpi_handle rootbus_handle; + unsigned long long segment; + acpi_status status; + struct acpi_buffer name_buffer = { ACPI_ALLOCATE_BUFFER, NULL }; + + rootbus_handle = acpi_device_handle(PCI_CONTROLLER(dev)->companion); + status = acpi_evaluate_integer(rootbus_handle, METHOD_NAME__SEG, NULL, + &segment); + if (ACPI_SUCCESS(status)) { + if (segment != pci_domain_nr(dev)) { + acpi_get_name(rootbus_handle, ACPI_FULL_PATHNAME, + &name_buffer); + printk(KERN_ERR + "%s: Segment number mismatch, 0x%llx vs 0x%x for: %s\n", + __func__, segment, pci_domain_nr(dev), + (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return 1; + } + } else { + acpi_get_name(rootbus_handle, ACPI_FULL_PATHNAME, &name_buffer); + printk(KERN_ERR "%s: Unable to get __SEG from: %s\n", + __func__, (char *)name_buffer.pointer); + kfree(name_buffer.pointer); + return 1; + } + + /* + * We want to search all devices in this segment/domain + * of the ACPI namespace for the matching ACPI device, + * which holds the pcidev_info pointer in its vendor resource. + */ + pcidev_match.bus = dev->bus->number; + pcidev_match.devfn = dev->devfn; + pcidev_match.handle = NULL; + + acpi_walk_namespace(ACPI_TYPE_DEVICE, rootbus_handle, ACPI_UINT32_MAX, + find_matching_device, NULL, &pcidev_match, NULL); + + if (!pcidev_match.handle) { + printk(KERN_ERR + "%s: Could not find matching ACPI device for %s.\n", + __func__, pci_name(dev)); + return 1; + } + + if (sn_extract_device_info(pcidev_match.handle, pcidev_info, sn_irq_info)) + return 1; + + /* Build up the pcidev_info.pdi_slot_host_handle */ + host_devfn = get_host_devfn(pcidev_match.handle, rootbus_handle); + (*pcidev_info)->pdi_slot_host_handle = + ((unsigned long) pci_domain_nr(dev) << 40) | + /* bus == 0 */ + host_devfn; + return 0; +} + +/* + * sn_acpi_slot_fixup - Obtain the pcidev_info and sn_irq_info. + * Perform any SN specific slot fixup. + * At present there does not appear to be + * any generic way to handle a ROM image + * that has been shadowed by the PROM, so + * we pass a pointer to it within the + * pcidev_info structure. + */ + +void +sn_acpi_slot_fixup(struct pci_dev *dev) +{ + void __iomem *addr; + struct pcidev_info *pcidev_info = NULL; + struct sn_irq_info *sn_irq_info = NULL; + size_t image_size, size; + + if (sn_acpi_get_pcidev_info(dev, &pcidev_info, &sn_irq_info)) { + panic("%s: Failure obtaining pcidev_info for %s\n", + __func__, pci_name(dev)); + } + + if (pcidev_info->pdi_pio_mapped_addr[PCI_ROM_RESOURCE]) { + /* + * A valid ROM image exists and has been shadowed by the + * PROM. Setup the pci_dev ROM resource with the address + * of the shadowed copy, and the actual length of the ROM image. + */ + size = pci_resource_len(dev, PCI_ROM_RESOURCE); + addr = ioremap(pcidev_info->pdi_pio_mapped_addr[PCI_ROM_RESOURCE], + size); + image_size = pci_get_rom_size(dev, addr, size); + dev->resource[PCI_ROM_RESOURCE].start = (unsigned long) addr; + dev->resource[PCI_ROM_RESOURCE].end = + (unsigned long) addr + image_size - 1; + dev->resource[PCI_ROM_RESOURCE].flags |= IORESOURCE_ROM_BIOS_COPY; + } + sn_pci_fixup_slot(dev, pcidev_info, sn_irq_info); +} + +EXPORT_SYMBOL(sn_acpi_slot_fixup); + + +/* + * sn_acpi_bus_fixup - Perform SN specific setup of software structs + * (pcibus_bussoft, pcidev_info) and hardware + * registers, for the specified bus and devices under it. + */ +void +sn_acpi_bus_fixup(struct pci_bus *bus) +{ + struct pci_dev *pci_dev = NULL; + struct pcibus_bussoft *prom_bussoft_ptr; + + if (!bus->parent) { /* If root bus */ + prom_bussoft_ptr = sn_get_bussoft_ptr(bus); + if (prom_bussoft_ptr == NULL) { + printk(KERN_ERR + "%s: 0x%04x:0x%02x Unable to " + "obtain prom_bussoft_ptr\n", + __func__, pci_domain_nr(bus), bus->number); + return; + } + sn_common_bus_fixup(bus, prom_bussoft_ptr); + } + list_for_each_entry(pci_dev, &bus->devices, bus_list) { + sn_acpi_slot_fixup(pci_dev); + } +} + +/* + * sn_io_acpi_init - PROM has ACPI support for IO, defining at a minimum the + * nodes and root buses in the DSDT. As a result, bus scanning + * will be initiated by the Linux ACPI code. + */ + +void __init +sn_io_acpi_init(void) +{ + u64 result; + long status; + + /* SN Altix does not follow the IOSAPIC IRQ routing model */ + acpi_irq_model = ACPI_IRQ_MODEL_PLATFORM; + + /* Setup hubdev_info for all SGIHUB/SGITIO devices */ + acpi_get_devices("SGIHUB", sn_acpi_hubdev_init, NULL, NULL); + acpi_get_devices("SGITIO", sn_acpi_hubdev_init, NULL, NULL); + + status = sal_ioif_init(&result); + if (status || result) + panic("sal_ioif_init failed: [%lx] %s\n", + status, ia64_sal_strerror(status)); +} diff --git a/arch/ia64/sn/kernel/io_common.c b/arch/ia64/sn/kernel/io_common.c new file mode 100644 index 00000000000..11f2275570f --- /dev/null +++ b/arch/ia64/sn/kernel/io_common.c @@ -0,0 +1,564 @@ +/* + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2006 Silicon Graphics, Inc. All rights reserved. + */ + +#include <linux/bootmem.h> +#include <linux/export.h> +#include <linux/slab.h> +#include <asm/sn/types.h> +#include <asm/sn/addrs.h> +#include <asm/sn/sn_feature_sets.h> +#include <asm/sn/geo.h> +#include <asm/sn/io.h> +#include <asm/sn/l1.h> +#include <asm/sn/module.h> +#include <asm/sn/pcibr_provider.h> +#include <asm/sn/pcibus_provider_defs.h> +#include <asm/sn/pcidev.h> +#include <asm/sn/simulator.h> +#include <asm/sn/sn_sal.h> +#include <asm/sn/tioca_provider.h> +#include <asm/sn/tioce_provider.h> +#include "xtalk/hubdev.h" +#include "xtalk/xwidgetdev.h" +#include <linux/acpi.h> +#include <asm/sn/sn2/sn_hwperf.h> +#include <asm/sn/acpi.h> + +extern void sn_init_cpei_timer(void); +extern void register_sn_procfs(void); +extern void sn_io_acpi_init(void); +extern void sn_io_init(void); + + +static struct list_head sn_sysdata_list; + +/* sysdata list struct */ +struct sysdata_el { + struct list_head entry; + void *sysdata; +}; + +int sn_ioif_inited; /* SN I/O infrastructure initialized? */ + +int sn_acpi_rev; /* SN ACPI revision */ +EXPORT_SYMBOL_GPL(sn_acpi_rev); + +struct sn_pcibus_provider *sn_pci_provider[PCIIO_ASIC_MAX_TYPES]; /* indexed by asic type */ + +/* + * Hooks and struct for unsupported pci providers + */ + +static dma_addr_t +sn_default_pci_map(struct pci_dev *pdev, unsigned long paddr, size_t size, int type) +{ + return 0; +} + +static void +sn_default_pci_unmap(struct pci_dev *pdev, dma_addr_t addr, int direction) +{ + return; +} + +static void * +sn_default_pci_bus_fixup(struct pcibus_bussoft *soft, struct pci_controller *controller) +{ + return NULL; +} + +static struct sn_pcibus_provider sn_pci_default_provider = { + .dma_map = sn_default_pci_map, + .dma_map_consistent = sn_default_pci_map, + .dma_unmap = sn_default_pci_unmap, + .bus_fixup = sn_default_pci_bus_fixup, +}; + +/* + * Retrieve the DMA Flush List given nasid, widget, and device. + * This list is needed to implement the WAR - Flush DMA data on PIO Reads. + */ +static inline u64 +sal_get_device_dmaflush_list(u64 nasid, u64 widget_num, u64 device_num, + u64 address) +{ + struct ia64_sal_retval ret_stuff; + ret_stuff.status = 0; + ret_stuff.v0 = 0; + + SAL_CALL_NOLOCK(ret_stuff, + (u64) SN_SAL_IOIF_GET_DEVICE_DMAFLUSH_LIST, + (u64) nasid, (u64) widget_num, + (u64) device_num, (u64) address, 0, 0, 0); + return ret_stuff.status; +} + +/* + * sn_pcidev_info_get() - Retrieve the pcidev_info struct for the specified + * device. + */ +inline struct pcidev_info * +sn_pcidev_info_get(struct pci_dev *dev) +{ + struct pcidev_info *pcidev; + + list_for_each_entry(pcidev, + &(SN_PLATFORM_DATA(dev)->pcidev_info), pdi_list) { + if (pcidev->pdi_linux_pcidev == dev) + return pcidev; + } + return NULL; +} + +/* Older PROM flush WAR + * + * 01/16/06 -- This war will be in place until a new official PROM is released. + * Additionally note that the struct sn_flush_device_war also has to be + * removed from arch/ia64/sn/include/xtalk/hubdev.h + */ + +static s64 sn_device_fixup_war(u64 nasid, u64 widget, int device, + struct sn_flush_device_common *common) +{ + struct sn_flush_device_war *war_list; + struct sn_flush_device_war *dev_entry; + struct ia64_sal_retval isrv = {0,0,0,0}; + + printk_once(KERN_WARNING + "PROM version < 4.50 -- implementing old PROM flush WAR\n"); + + war_list = kzalloc(DEV_PER_WIDGET * sizeof(*war_list), GFP_KERNEL); + BUG_ON(!war_list); + + SAL_CALL_NOLOCK(isrv, SN_SAL_IOIF_GET_WIDGET_DMAFLUSH_LIST, + nasid, widget, __pa(war_list), 0, 0, 0 ,0); + if (isrv.status) + panic("sn_device_fixup_war failed: %s\n", + ia64_sal_strerror(isrv.status)); + + dev_entry = war_list + device; + memcpy(common,dev_entry, sizeof(*common)); + kfree(war_list); + + return isrv.status; +} + +/* + * sn_common_hubdev_init() - This routine is called to initialize the HUB data + * structure for each node in the system. + */ +void __init +sn_common_hubdev_init(struct hubdev_info *hubdev) +{ + + struct sn_flush_device_kernel *sn_flush_device_kernel; + struct sn_flush_device_kernel *dev_entry; + s64 status; + int widget, device, size; + + /* Attach the error interrupt handlers */ + if (hubdev->hdi_nasid & 1) /* If TIO */ + ice_error_init(hubdev); + else + hub_error_init(hubdev); + + for (widget = 0; widget <= HUB_WIDGET_ID_MAX; widget++) + hubdev->hdi_xwidget_info[widget].xwi_hubinfo = hubdev; + + if (!hubdev->hdi_flush_nasid_list.widget_p) + return; + + size = (HUB_WIDGET_ID_MAX + 1) * + sizeof(struct sn_flush_device_kernel *); + hubdev->hdi_flush_nasid_list.widget_p = + kzalloc(size, GFP_KERNEL); + BUG_ON(!hubdev->hdi_flush_nasid_list.widget_p); + + for (widget = 0; widget <= HUB_WIDGET_ID_MAX; widget++) { + size = DEV_PER_WIDGET * + sizeof(struct sn_flush_device_kernel); + sn_flush_device_kernel = kzalloc(size, GFP_KERNEL); + BUG_ON(!sn_flush_device_kernel); + + dev_entry = sn_flush_device_kernel; + for (device = 0; device < DEV_PER_WIDGET; + device++, dev_entry++) { + size = sizeof(struct sn_flush_device_common); + dev_entry->common = kzalloc(size, GFP_KERNEL); + BUG_ON(!dev_entry->common); + if (sn_prom_feature_available(PRF_DEVICE_FLUSH_LIST)) + status = sal_get_device_dmaflush_list( + hubdev->hdi_nasid, widget, device, + (u64)(dev_entry->common)); + else + status = sn_device_fixup_war(hubdev->hdi_nasid, + widget, device, + dev_entry->common); + if (status != SALRET_OK) + panic("SAL call failed: %s\n", + ia64_sal_strerror(status)); + + spin_lock_init(&dev_entry->sfdl_flush_lock); + } + + if (sn_flush_device_kernel) + hubdev->hdi_flush_nasid_list.widget_p[widget] = + sn_flush_device_kernel; + } +} + +void sn_pci_unfixup_slot(struct pci_dev *dev) +{ + struct pci_dev *host_pci_dev = SN_PCIDEV_INFO(dev)->host_pci_dev; + + sn_irq_unfixup(dev); + pci_dev_put(host_pci_dev); + pci_dev_put(dev); +} + +/* + * sn_pci_fixup_slot() + */ +void sn_pci_fixup_slot(struct pci_dev *dev, struct pcidev_info *pcidev_info, + struct sn_irq_info *sn_irq_info) +{ + int segment = pci_domain_nr(dev->bus); + struct pcibus_bussoft *bs; + struct pci_dev *host_pci_dev; + unsigned int bus_no, devfn; + + pci_dev_get(dev); /* for the sysdata pointer */ + + /* Add pcidev_info to list in pci_controller.platform_data */ + list_add_tail(&pcidev_info->pdi_list, + &(SN_PLATFORM_DATA(dev->bus)->pcidev_info)); + /* + * Using the PROMs values for the PCI host bus, get the Linux + * PCI host_pci_dev struct and set up host bus linkages + */ + + bus_no = (pcidev_info->pdi_slot_host_handle >> 32) & 0xff; + devfn = pcidev_info->pdi_slot_host_handle & 0xffffffff; + host_pci_dev = pci_get_domain_bus_and_slot(segment, bus_no, devfn); + + pcidev_info->host_pci_dev = host_pci_dev; + pcidev_info->pdi_linux_pcidev = dev; + pcidev_info->pdi_host_pcidev_info = SN_PCIDEV_INFO(host_pci_dev); + bs = SN_PCIBUS_BUSSOFT(dev->bus); + pcidev_info->pdi_pcibus_info = bs; + + if (bs && bs->bs_asic_type < PCIIO_ASIC_MAX_TYPES) { + SN_PCIDEV_BUSPROVIDER(dev) = sn_pci_provider[bs->bs_asic_type]; + } else { + SN_PCIDEV_BUSPROVIDER(dev) = &sn_pci_default_provider; + } + + /* Only set up IRQ stuff if this device has a host bus context */ + if (bs && sn_irq_info->irq_irq) { + pcidev_info->pdi_sn_irq_info = sn_irq_info; + dev->irq = pcidev_info->pdi_sn_irq_info->irq_irq; + sn_irq_fixup(dev, sn_irq_info); + } else { + pcidev_info->pdi_sn_irq_info = NULL; + kfree(sn_irq_info); + } +} + +/* + * sn_common_bus_fixup - Perform platform specific bus fixup. + * Execute the ASIC specific fixup routine + * for this bus. + */ +void +sn_common_bus_fixup(struct pci_bus *bus, + struct pcibus_bussoft *prom_bussoft_ptr) +{ + int cnode; + struct pci_controller *controller; + struct hubdev_info *hubdev_info; + int nasid; + void *provider_soft; + struct sn_pcibus_provider *provider; + struct sn_platform_data *sn_platform_data; + + controller = PCI_CONTROLLER(bus); + /* + * Per-provider fixup. Copies the bus soft structure from prom + * to local area and links SN_PCIBUS_BUSSOFT(). + */ + + if (prom_bussoft_ptr->bs_asic_type >= PCIIO_ASIC_MAX_TYPES) { + printk(KERN_WARNING "sn_common_bus_fixup: Unsupported asic type, %d", + prom_bussoft_ptr->bs_asic_type); + return; + } + + if (prom_bussoft_ptr->bs_asic_type == PCIIO_ASIC_TYPE_PPB) + return; /* no further fixup necessary */ + + provider = sn_pci_provider[prom_bussoft_ptr->bs_asic_type]; + if (provider == NULL) + panic("sn_common_bus_fixup: No provider registered for this asic type, %d", + prom_bussoft_ptr->bs_asic_type); + + if (provider->bus_fixup) + provider_soft = (*provider->bus_fixup) (prom_bussoft_ptr, + controller); + else + provider_soft = NULL; + + /* + * Generic bus fixup goes here. Don't reference prom_bussoft_ptr + * after this point. + */ + controller->platform_data = kzalloc(sizeof(struct sn_platform_data), + GFP_KERNEL); + BUG_ON(controller->platform_data == NULL); + sn_platform_data = + (struct sn_platform_data *) controller->platform_data; + sn_platform_data->provider_soft = provider_soft; + INIT_LIST_HEAD(&((struct sn_platform_data *) + controller->platform_data)->pcidev_info); + nasid = NASID_GET(SN_PCIBUS_BUSSOFT(bus)->bs_base); + cnode = nasid_to_cnodeid(nasid); + hubdev_info = (struct hubdev_info *)(NODEPDA(cnode)->pdinfo); + SN_PCIBUS_BUSSOFT(bus)->bs_xwidget_info = + &(hubdev_info->hdi_xwidget_info[SN_PCIBUS_BUSSOFT(bus)->bs_xid]); + + /* + * If the node information we obtained during the fixup phase is + * invalid then set controller->node to -1 (undetermined) + */ + if (controller->node >= num_online_nodes()) { + struct pcibus_bussoft *b = SN_PCIBUS_BUSSOFT(bus); + + printk(KERN_WARNING "Device ASIC=%u XID=%u PBUSNUM=%u " + "L_IO=%llx L_MEM=%llx BASE=%llx\n", + b->bs_asic_type, b->bs_xid, b->bs_persist_busnum, + b->bs_legacy_io, b->bs_legacy_mem, b->bs_base); + printk(KERN_WARNING "on node %d but only %d nodes online." + "Association set to undetermined.\n", + controller->node, num_online_nodes()); + controller->node = -1; + } +} + +void sn_bus_store_sysdata(struct pci_dev *dev) +{ + struct sysdata_el *element; + + element = kzalloc(sizeof(struct sysdata_el), GFP_KERNEL); + if (!element) { + dev_dbg(&dev->dev, "%s: out of memory!\n", __func__); + return; + } + element->sysdata = SN_PCIDEV_INFO(dev); + list_add(&element->entry, &sn_sysdata_list); +} + +void sn_bus_free_sysdata(void) +{ + struct sysdata_el *element; + struct list_head *list, *safe; + + list_for_each_safe(list, safe, &sn_sysdata_list) { + element = list_entry(list, struct sysdata_el, entry); + list_del(&element->entry); + list_del(&(((struct pcidev_info *) + (element->sysdata))->pdi_list)); + kfree(element->sysdata); + kfree(element); + } + return; +} + +/* + * hubdev_init_node() - Creates the HUB data structure and link them to it's + * own NODE specific data area. + */ +void __init hubdev_init_node(nodepda_t * npda, cnodeid_t node) +{ + struct hubdev_info *hubdev_info; + int size; + pg_data_t *pg; + + size = sizeof(struct hubdev_info); + + if (node >= num_online_nodes()) /* Headless/memless IO nodes */ + pg = NODE_DATA(0); + else + pg = NODE_DATA(node); + + hubdev_info = (struct hubdev_info *)alloc_bootmem_node(pg, size); + + npda->pdinfo = (void *)hubdev_info; +} + +geoid_t +cnodeid_get_geoid(cnodeid_t cnode) +{ + struct hubdev_info *hubdev; + + hubdev = (struct hubdev_info *)(NODEPDA(cnode)->pdinfo); + return hubdev->hdi_geoid; +} + +void sn_generate_path(struct pci_bus *pci_bus, char *address) +{ + nasid_t nasid; + cnodeid_t cnode; + geoid_t geoid; + moduleid_t moduleid; + u16 bricktype; + + nasid = NASID_GET(SN_PCIBUS_BUSSOFT(pci_bus)->bs_base); + cnode = nasid_to_cnodeid(nasid); + geoid = cnodeid_get_geoid(cnode); + moduleid = geo_module(geoid); + + sprintf(address, "module_%c%c%c%c%.2d", + '0'+RACK_GET_CLASS(MODULE_GET_RACK(moduleid)), + '0'+RACK_GET_GROUP(MODULE_GET_RACK(moduleid)), + '0'+RACK_GET_NUM(MODULE_GET_RACK(moduleid)), + MODULE_GET_BTCHAR(moduleid), MODULE_GET_BPOS(moduleid)); + + /* Tollhouse requires slot id to be displayed */ + bricktype = MODULE_GET_BTYPE(moduleid); + if ((bricktype == L1_BRICKTYPE_191010) || + (bricktype == L1_BRICKTYPE_1932)) + sprintf(address + strlen(address), "^%d", + geo_slot(geoid)); +} + +void sn_pci_fixup_bus(struct pci_bus *bus) +{ + + if (SN_ACPI_BASE_SUPPORT()) + sn_acpi_bus_fixup(bus); + else + sn_bus_fixup(bus); +} + +/* + * sn_io_early_init - Perform early IO (and some non-IO) initialization. + * In particular, setup the sn_pci_provider[] array. + * This needs to be done prior to any bus scanning + * (acpi_scan_init()) in the ACPI case, as the SN + * bus fixup code will reference the array. + */ +static int __init +sn_io_early_init(void) +{ + int i; + + if (!ia64_platform_is("sn2") || IS_RUNNING_ON_FAKE_PROM()) + return 0; + + /* we set the acpi revision to that of the DSDT table OEM rev. */ + { + struct acpi_table_header *header = NULL; + + acpi_get_table(ACPI_SIG_DSDT, 1, &header); + BUG_ON(header == NULL); + sn_acpi_rev = header->oem_revision; + } + + /* + * prime sn_pci_provider[]. Individual provider init routines will + * override their respective default entries. + */ + + for (i = 0; i < PCIIO_ASIC_MAX_TYPES; i++) + sn_pci_provider[i] = &sn_pci_default_provider; + + pcibr_init_provider(); + tioca_init_provider(); + tioce_init_provider(); + + /* + * This is needed to avoid bounce limit checks in the blk layer + */ + ia64_max_iommu_merge_mask = ~PAGE_MASK; + + sn_irq_lh_init(); + INIT_LIST_HEAD(&sn_sysdata_list); + sn_init_cpei_timer(); + +#ifdef CONFIG_PROC_FS + register_sn_procfs(); +#endif + + { + struct acpi_table_header *header; + (void)acpi_get_table(ACPI_SIG_DSDT, 1, &header); + printk(KERN_INFO "ACPI DSDT OEM Rev 0x%x\n", + header->oem_revision); + } + if (SN_ACPI_BASE_SUPPORT()) + sn_io_acpi_init(); + else + sn_io_init(); + return 0; +} + +arch_initcall(sn_io_early_init); + +/* + * sn_io_late_init() - Perform any final platform specific IO initialization. + */ + +int __init +sn_io_late_init(void) +{ + struct pci_bus *bus; + struct pcibus_bussoft *bussoft; + cnodeid_t cnode; + nasid_t nasid; + cnodeid_t near_cnode; + + if (!ia64_platform_is("sn2") || IS_RUNNING_ON_FAKE_PROM()) + return 0; + + /* + * Setup closest node in pci_controller->node for + * PIC, TIOCP, TIOCE (TIOCA does it during bus fixup using + * info from the PROM). + */ + bus = NULL; + while ((bus = pci_find_next_bus(bus)) != NULL) { + bussoft = SN_PCIBUS_BUSSOFT(bus); + nasid = NASID_GET(bussoft->bs_base); + cnode = nasid_to_cnodeid(nasid); + if ((bussoft->bs_asic_type == PCIIO_ASIC_TYPE_TIOCP) || + (bussoft->bs_asic_type == PCIIO_ASIC_TYPE_TIOCE) || + (bussoft->bs_asic_type == PCIIO_ASIC_TYPE_PIC)) { + /* PCI Bridge: find nearest node with CPUs */ + int e = sn_hwperf_get_nearest_node(cnode, NULL, + &near_cnode); + if (e < 0) { + near_cnode = (cnodeid_t)-1; /* use any node */ + printk(KERN_WARNING "sn_io_late_init: failed " + "to find near node with CPUs for " + "node %d, err=%d\n", cnode, e); + } + PCI_CONTROLLER(bus)->node = near_cnode; + } + } + + sn_ioif_inited = 1; /* SN I/O infrastructure now initialized */ + + return 0; +} + +fs_initcall(sn_io_late_init); + +EXPORT_SYMBOL(sn_pci_unfixup_slot); +EXPORT_SYMBOL(sn_bus_store_sysdata); +EXPORT_SYMBOL(sn_bus_free_sysdata); +EXPORT_SYMBOL(sn_generate_path); + diff --git a/arch/ia64/sn/kernel/io_init.c b/arch/ia64/sn/kernel/io_init.c index dc09a6a28a3..0b5ce82d203 100644 --- a/arch/ia64/sn/kernel/io_init.c +++ b/arch/ia64/sn/kernel/io_init.c @@ -3,103 +3,30 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 1992 - 1997, 2000-2005 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 1992 - 1997, 2000-2006 Silicon Graphics, Inc. All rights reserved. */ -#include <linux/bootmem.h> -#include <linux/nodemask.h> +#include <linux/slab.h> +#include <linux/export.h> #include <asm/sn/types.h> #include <asm/sn/addrs.h> -#include <asm/sn/sn_feature_sets.h> -#include <asm/sn/geo.h> #include <asm/sn/io.h> -#include <asm/sn/l1.h> #include <asm/sn/module.h> -#include <asm/sn/pcibr_provider.h> +#include <asm/sn/intr.h> #include <asm/sn/pcibus_provider_defs.h> #include <asm/sn/pcidev.h> -#include <asm/sn/simulator.h> #include <asm/sn/sn_sal.h> -#include <asm/sn/tioca_provider.h> -#include <asm/sn/tioce_provider.h> #include "xtalk/hubdev.h" -#include "xtalk/xwidgetdev.h" - - -extern void sn_init_cpei_timer(void); -extern void register_sn_procfs(void); - -static struct list_head sn_sysdata_list; - -/* sysdata list struct */ -struct sysdata_el { - struct list_head entry; - void *sysdata; -}; - -struct slab_info { - struct hubdev_info hubdev; -}; - -struct brick { - moduleid_t id; /* Module ID of this module */ - struct slab_info slab_info[MAX_SLABS + 1]; -}; - -int sn_ioif_inited; /* SN I/O infrastructure initialized? */ - -struct sn_pcibus_provider *sn_pci_provider[PCIIO_ASIC_MAX_TYPES]; /* indexed by asic type */ - -static int max_segment_number; /* Default highest segment number */ -static int max_pcibus_number = 255; /* Default highest pci bus number */ /* - * Hooks and struct for unsupported pci providers + * The code in this file will only be executed when running with + * a PROM that does _not_ have base ACPI IO support. + * (i.e., SN_ACPI_BASE_SUPPORT() == 0) */ -static dma_addr_t -sn_default_pci_map(struct pci_dev *pdev, unsigned long paddr, size_t size, int type) -{ - return 0; -} - -static void -sn_default_pci_unmap(struct pci_dev *pdev, dma_addr_t addr, int direction) -{ - return; -} - -static void * -sn_default_pci_bus_fixup(struct pcibus_bussoft *soft, struct pci_controller *controller) -{ - return NULL; -} - -static struct sn_pcibus_provider sn_pci_default_provider = { - .dma_map = sn_default_pci_map, - .dma_map_consistent = sn_default_pci_map, - .dma_unmap = sn_default_pci_unmap, - .bus_fixup = sn_default_pci_bus_fixup, -}; - -/* - * Retrieve the DMA Flush List given nasid, widget, and device. - * This list is needed to implement the WAR - Flush DMA data on PIO Reads. - */ -static inline u64 -sal_get_device_dmaflush_list(u64 nasid, u64 widget_num, u64 device_num, - u64 address) -{ - struct ia64_sal_retval ret_stuff; - ret_stuff.status = 0; - ret_stuff.v0 = 0; +static int max_segment_number; /* Default highest segment number */ +static int max_pcibus_number = 255; /* Default highest pci bus number */ - SAL_CALL_NOLOCK(ret_stuff, - (u64) SN_SAL_IOIF_GET_DEVICE_DMAFLUSH_LIST, - (u64) nasid, (u64) widget_num, - (u64) device_num, (u64) address, 0, 0, 0); - return ret_stuff.status; -} /* * Retrieve the hub device info structure for the given nasid. @@ -150,74 +77,20 @@ sal_get_pcidev_info(u64 segment, u64 bus_number, u64 devfn, u64 pci_dev, return ret_stuff.v0; } -/* - * sn_pcidev_info_get() - Retrieve the pcidev_info struct for the specified - * device. - */ -inline struct pcidev_info * -sn_pcidev_info_get(struct pci_dev *dev) -{ - struct pcidev_info *pcidev; - - list_for_each_entry(pcidev, - &(SN_PCI_CONTROLLER(dev)->pcidev_info), pdi_list) { - if (pcidev->pdi_linux_pcidev == dev) { - return pcidev; - } - } - return NULL; -} - -/* Older PROM flush WAR - * - * 01/16/06 -- This war will be in place until a new official PROM is released. - * Additionally note that the struct sn_flush_device_war also has to be - * removed from arch/ia64/sn/include/xtalk/hubdev.h - */ -static u8 war_implemented = 0; - -static s64 sn_device_fixup_war(u64 nasid, u64 widget, int device, - struct sn_flush_device_common *common) -{ - struct sn_flush_device_war *war_list; - struct sn_flush_device_war *dev_entry; - struct ia64_sal_retval isrv = {0,0,0,0}; - - if (!war_implemented) { - printk(KERN_WARNING "PROM version < 4.50 -- implementing old " - "PROM flush WAR\n"); - war_implemented = 1; - } - - war_list = kzalloc(DEV_PER_WIDGET * sizeof(*war_list), GFP_KERNEL); - if (!war_list) - BUG(); - - SAL_CALL_NOLOCK(isrv, SN_SAL_IOIF_GET_WIDGET_DMAFLUSH_LIST, - nasid, widget, __pa(war_list), 0, 0, 0 ,0); - if (isrv.status) - panic("sn_device_fixup_war failed: %s\n", - ia64_sal_strerror(isrv.status)); - - dev_entry = war_list + device; - memcpy(common,dev_entry, sizeof(*common)); - kfree(war_list); - - return isrv.status; -} /* - * sn_fixup_ionodes() - This routine initializes the HUB data strcuture for - * each node in the system. + * sn_fixup_ionodes() - This routine initializes the HUB data structure for + * each node in the system. This function is only + * executed when running with a non-ACPI capable PROM. */ static void __init sn_fixup_ionodes(void) { - struct sn_flush_device_kernel *sn_flush_device_kernel; - struct sn_flush_device_kernel *dev_entry; + struct hubdev_info *hubdev; u64 status; u64 nasid; - int i, widget, device, size; + int i; + extern void sn_common_hubdev_init(struct hubdev_info *); /* * Get SGI Specific HUB chipset information. @@ -240,165 +113,70 @@ static void __init sn_fixup_ionodes(void) max_segment_number = hubdev->max_segment_number; max_pcibus_number = hubdev->max_pcibus_number; } - - /* Attach the error interrupt handlers */ - if (nasid & 1) - ice_error_init(hubdev); - else - hub_error_init(hubdev); - - for (widget = 0; widget <= HUB_WIDGET_ID_MAX; widget++) - hubdev->hdi_xwidget_info[widget].xwi_hubinfo = hubdev; - - if (!hubdev->hdi_flush_nasid_list.widget_p) - continue; - - size = (HUB_WIDGET_ID_MAX + 1) * - sizeof(struct sn_flush_device_kernel *); - hubdev->hdi_flush_nasid_list.widget_p = - kzalloc(size, GFP_KERNEL); - if (!hubdev->hdi_flush_nasid_list.widget_p) - BUG(); - - for (widget = 0; widget <= HUB_WIDGET_ID_MAX; widget++) { - size = DEV_PER_WIDGET * - sizeof(struct sn_flush_device_kernel); - sn_flush_device_kernel = kzalloc(size, GFP_KERNEL); - if (!sn_flush_device_kernel) - BUG(); - - dev_entry = sn_flush_device_kernel; - for (device = 0; device < DEV_PER_WIDGET; - device++,dev_entry++) { - size = sizeof(struct sn_flush_device_common); - dev_entry->common = kzalloc(size, GFP_KERNEL); - if (!dev_entry->common) - BUG(); - - if (sn_prom_feature_available( - PRF_DEVICE_FLUSH_LIST)) - status = sal_get_device_dmaflush_list( - nasid, widget, device, - (u64)(dev_entry->common)); - else - status = sn_device_fixup_war(nasid, - widget, device, - dev_entry->common); - if (status != SALRET_OK) - panic("SAL call failed: %s\n", - ia64_sal_strerror(status)); - - spin_lock_init(&dev_entry->sfdl_flush_lock); - } - - if (sn_flush_device_kernel) - hubdev->hdi_flush_nasid_list.widget_p[widget] = - sn_flush_device_kernel; - } + sn_common_hubdev_init(hubdev); } } /* - * sn_pci_window_fixup() - Create a pci_window for each device resource. - * Until ACPI support is added, we need this code - * to setup pci_windows for use by - * pcibios_bus_to_resource(), - * pcibios_resource_to_bus(), etc. + * sn_pci_legacy_window_fixup - Setup PCI resources for + * legacy IO and MEM space. This needs to + * be done here, as the PROM does not have + * ACPI support defining the root buses + * and their resources (_CRS), */ static void -sn_pci_window_fixup(struct pci_dev *dev, unsigned int count, - s64 * pci_addrs) -{ - struct pci_controller *controller = PCI_CONTROLLER(dev->bus); - unsigned int i; - unsigned int idx; - unsigned int new_count; - struct pci_window *new_window; - - if (count == 0) - return; - idx = controller->windows; - new_count = controller->windows + count; - new_window = kcalloc(new_count, sizeof(struct pci_window), GFP_KERNEL); - if (new_window == NULL) - BUG(); - if (controller->window) { - memcpy(new_window, controller->window, - sizeof(struct pci_window) * controller->windows); - kfree(controller->window); - } - - /* Setup a pci_window for each device resource. */ - for (i = 0; i <= PCI_ROM_RESOURCE; i++) { - if (pci_addrs[i] == -1) - continue; - - new_window[idx].offset = dev->resource[i].start - pci_addrs[i]; - new_window[idx].resource = dev->resource[i]; - idx++; - } - - controller->windows = new_count; - controller->window = new_window; -} - -void sn_pci_unfixup_slot(struct pci_dev *dev) +sn_legacy_pci_window_fixup(struct resource *res, + u64 legacy_io, u64 legacy_mem) { - struct pci_dev *host_pci_dev = SN_PCIDEV_INFO(dev)->host_pci_dev; - - sn_irq_unfixup(dev); - pci_dev_put(host_pci_dev); - pci_dev_put(dev); + res[0].name = "legacy_io"; + res[0].flags = IORESOURCE_IO; + res[0].start = legacy_io; + res[0].end = res[0].start + 0xffff; + res[0].parent = &ioport_resource; + res[1].name = "legacy_mem"; + res[1].flags = IORESOURCE_MEM; + res[1].start = legacy_mem; + res[1].end = res[1].start + (1024 * 1024) - 1; + res[1].parent = &iomem_resource; } /* - * sn_pci_fixup_slot() - This routine sets up a slot's resources - * consistent with the Linux PCI abstraction layer. Resources acquired - * from our PCI provider include PIO maps to BAR space and interrupt - * objects. + * sn_io_slot_fixup() - We are not running with an ACPI capable PROM, + * and need to convert the pci_dev->resource + * 'start' and 'end' addresses to mapped addresses, + * and setup the pci_controller->window array entries. */ -void sn_pci_fixup_slot(struct pci_dev *dev) +void +sn_io_slot_fixup(struct pci_dev *dev) { - unsigned int count = 0; int idx; - int segment = pci_domain_nr(dev->bus); - int status = 0; - struct pcibus_bussoft *bs; - struct pci_bus *host_pci_bus; - struct pci_dev *host_pci_dev; + unsigned long addr, end, size, start; struct pcidev_info *pcidev_info; - s64 pci_addrs[PCI_ROM_RESOURCE + 1]; - struct sn_irq_info *sn_irq_info; - unsigned long size; - unsigned int bus_no, devfn; + struct sn_irq_info *sn_irq_info; + int status; - pci_dev_get(dev); /* for the sysdata pointer */ pcidev_info = kzalloc(sizeof(struct pcidev_info), GFP_KERNEL); if (!pcidev_info) - BUG(); /* Cannot afford to run out of memory */ + panic("%s: Unable to alloc memory for pcidev_info", __func__); sn_irq_info = kzalloc(sizeof(struct sn_irq_info), GFP_KERNEL); if (!sn_irq_info) - BUG(); /* Cannot afford to run out of memory */ + panic("%s: Unable to alloc memory for sn_irq_info", __func__); /* Call to retrieve pci device information needed by kernel. */ - status = sal_get_pcidev_info((u64) segment, (u64) dev->bus->number, - dev->devfn, - (u64) __pa(pcidev_info), - (u64) __pa(sn_irq_info)); - if (status) - BUG(); /* Cannot get platform pci device information */ + status = sal_get_pcidev_info((u64) pci_domain_nr(dev), + (u64) dev->bus->number, + dev->devfn, + (u64) __pa(pcidev_info), + (u64) __pa(sn_irq_info)); + + BUG_ON(status); /* Cannot get platform pci device information */ - /* Add pcidev_info to list in sn_pci_controller struct */ - list_add_tail(&pcidev_info->pdi_list, - &(SN_PCI_CONTROLLER(dev->bus)->pcidev_info)); /* Copy over PIO Mapped Addresses */ for (idx = 0; idx <= PCI_ROM_RESOURCE; idx++) { - unsigned long start, end, addr; if (!pcidev_info->pdi_pio_mapped_addr[idx]) { - pci_addrs[idx] = -1; continue; } @@ -406,73 +184,60 @@ void sn_pci_fixup_slot(struct pci_dev *dev) end = dev->resource[idx].end; size = end - start; if (size == 0) { - pci_addrs[idx] = -1; continue; } - pci_addrs[idx] = start; - count++; addr = pcidev_info->pdi_pio_mapped_addr[idx]; addr = ((addr << 4) >> 4) | __IA64_UNCACHED_OFFSET; dev->resource[idx].start = addr; dev->resource[idx].end = addr + size; + + /* + * if it's already in the device structure, remove it before + * inserting + */ + if (dev->resource[idx].parent && dev->resource[idx].parent->child) + release_resource(&dev->resource[idx]); + if (dev->resource[idx].flags & IORESOURCE_IO) - dev->resource[idx].parent = &ioport_resource; + insert_resource(&ioport_resource, &dev->resource[idx]); else - dev->resource[idx].parent = &iomem_resource; - } - /* Create a pci_window in the pci_controller struct for - * each device resource. - */ - if (count > 0) - sn_pci_window_fixup(dev, count, pci_addrs); - - /* - * Using the PROMs values for the PCI host bus, get the Linux - * PCI host_pci_dev struct and set up host bus linkages - */ - - bus_no = (pcidev_info->pdi_slot_host_handle >> 32) & 0xff; - devfn = pcidev_info->pdi_slot_host_handle & 0xffffffff; - host_pci_bus = pci_find_bus(segment, bus_no); - host_pci_dev = pci_get_slot(host_pci_bus, devfn); - - pcidev_info->host_pci_dev = host_pci_dev; - pcidev_info->pdi_linux_pcidev = dev; - pcidev_info->pdi_host_pcidev_info = SN_PCIDEV_INFO(host_pci_dev); - bs = SN_PCIBUS_BUSSOFT(dev->bus); - pcidev_info->pdi_pcibus_info = bs; - - if (bs && bs->bs_asic_type < PCIIO_ASIC_MAX_TYPES) { - SN_PCIDEV_BUSPROVIDER(dev) = sn_pci_provider[bs->bs_asic_type]; - } else { - SN_PCIDEV_BUSPROVIDER(dev) = &sn_pci_default_provider; + insert_resource(&iomem_resource, &dev->resource[idx]); + /* + * If ROM, set the actual ROM image size, and mark as + * shadowed in PROM. + */ + if (idx == PCI_ROM_RESOURCE) { + size_t image_size; + void __iomem *rom; + + rom = ioremap(pci_resource_start(dev, PCI_ROM_RESOURCE), + size + 1); + image_size = pci_get_rom_size(dev, rom, size + 1); + dev->resource[PCI_ROM_RESOURCE].end = + dev->resource[PCI_ROM_RESOURCE].start + + image_size - 1; + dev->resource[PCI_ROM_RESOURCE].flags |= + IORESOURCE_ROM_BIOS_COPY; + } } - /* Only set up IRQ stuff if this device has a host bus context */ - if (bs && sn_irq_info->irq_irq) { - pcidev_info->pdi_sn_irq_info = sn_irq_info; - dev->irq = pcidev_info->pdi_sn_irq_info->irq_irq; - sn_irq_fixup(dev, sn_irq_info); - } else { - pcidev_info->pdi_sn_irq_info = NULL; - kfree(sn_irq_info); - } + sn_pci_fixup_slot(dev, pcidev_info, sn_irq_info); } +EXPORT_SYMBOL(sn_io_slot_fixup); + /* * sn_pci_controller_fixup() - This routine sets up a bus's resources - * consistent with the Linux PCI abstraction layer. + * consistent with the Linux PCI abstraction layer. */ -void sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus) +static void __init +sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus) { - int status; - int nasid, cnode; + s64 status = 0; struct pci_controller *controller; - struct sn_pci_controller *sn_controller; struct pcibus_bussoft *prom_bussoft_ptr; - struct hubdev_info *hubdev_info; - void *provider_soft; - struct sn_pcibus_provider *provider; + struct resource *res; + LIST_HEAD(resources); status = sal_get_pcibus_info((u64) segment, (u64) busnum, (u64) ia64_tpa(&prom_bussoft_ptr)); @@ -480,261 +245,75 @@ void sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus) return; /*bus # does not exist */ prom_bussoft_ptr = __va(prom_bussoft_ptr); - /* Allocate a sn_pci_controller, which has a pci_controller struct - * as the first member. - */ - sn_controller = kzalloc(sizeof(struct sn_pci_controller), GFP_KERNEL); - if (!sn_controller) - BUG(); - INIT_LIST_HEAD(&sn_controller->pcidev_info); - controller = &sn_controller->pci_controller; + controller = kzalloc(sizeof(*controller), GFP_KERNEL); + BUG_ON(!controller); controller->segment = segment; - if (bus == NULL) { - bus = pci_scan_bus(busnum, &pci_root_ops, controller); - if (bus == NULL) - goto error_return; /* error, or bus already scanned */ - bus->sysdata = NULL; - } - - if (bus->sysdata) - goto error_return; /* sysdata already alloc'd */ - - /* - * Per-provider fixup. Copies the contents from prom to local - * area and links SN_PCIBUS_BUSSOFT(). - */ - - if (prom_bussoft_ptr->bs_asic_type >= PCIIO_ASIC_MAX_TYPES) - goto error_return; /* unsupported asic type */ - - if (prom_bussoft_ptr->bs_asic_type == PCIIO_ASIC_TYPE_PPB) - goto error_return; /* no further fixup necessary */ - - provider = sn_pci_provider[prom_bussoft_ptr->bs_asic_type]; - if (provider == NULL) - goto error_return; /* no provider registerd for this asic */ - - bus->sysdata = controller; - if (provider->bus_fixup) - provider_soft = (*provider->bus_fixup) (prom_bussoft_ptr, controller); - else - provider_soft = NULL; - - if (provider_soft == NULL) { - /* fixup failed or not applicable */ - bus->sysdata = NULL; - goto error_return; - } + res = kcalloc(2, sizeof(struct resource), GFP_KERNEL); + BUG_ON(!res); /* - * Setup pci_windows for legacy IO and MEM space. - * (Temporary until ACPI support is in place.) + * Temporarily save the prom_bussoft_ptr for use by sn_bus_fixup(). + * (platform_data will be overwritten later in sn_common_bus_fixup()) */ - controller->window = kcalloc(2, sizeof(struct pci_window), GFP_KERNEL); - if (controller->window == NULL) - BUG(); - controller->window[0].offset = prom_bussoft_ptr->bs_legacy_io; - controller->window[0].resource.name = "legacy_io"; - controller->window[0].resource.flags = IORESOURCE_IO; - controller->window[0].resource.start = prom_bussoft_ptr->bs_legacy_io; - controller->window[0].resource.end = - controller->window[0].resource.start + 0xffff; - controller->window[0].resource.parent = &ioport_resource; - controller->window[1].offset = prom_bussoft_ptr->bs_legacy_mem; - controller->window[1].resource.name = "legacy_mem"; - controller->window[1].resource.flags = IORESOURCE_MEM; - controller->window[1].resource.start = prom_bussoft_ptr->bs_legacy_mem; - controller->window[1].resource.end = - controller->window[1].resource.start + (1024 * 1024) - 1; - controller->window[1].resource.parent = &iomem_resource; - controller->windows = 2; - - /* - * Generic bus fixup goes here. Don't reference prom_bussoft_ptr - * after this point. - */ - - PCI_CONTROLLER(bus)->platform_data = provider_soft; - nasid = NASID_GET(SN_PCIBUS_BUSSOFT(bus)->bs_base); - cnode = nasid_to_cnodeid(nasid); - hubdev_info = (struct hubdev_info *)(NODEPDA(cnode)->pdinfo); - SN_PCIBUS_BUSSOFT(bus)->bs_xwidget_info = - &(hubdev_info->hdi_xwidget_info[SN_PCIBUS_BUSSOFT(bus)->bs_xid]); - - /* - * If the node information we obtained during the fixup phase is invalid - * then set controller->node to -1 (undetermined) - */ - if (controller->node >= num_online_nodes()) { - struct pcibus_bussoft *b = SN_PCIBUS_BUSSOFT(bus); - - printk(KERN_WARNING "Device ASIC=%u XID=%u PBUSNUM=%u" - "L_IO=%lx L_MEM=%lx BASE=%lx\n", - b->bs_asic_type, b->bs_xid, b->bs_persist_busnum, - b->bs_legacy_io, b->bs_legacy_mem, b->bs_base); - printk(KERN_WARNING "on node %d but only %d nodes online." - "Association set to undetermined.\n", - controller->node, num_online_nodes()); - controller->node = -1; + controller->platform_data = prom_bussoft_ptr; + + sn_legacy_pci_window_fixup(res, + prom_bussoft_ptr->bs_legacy_io, + prom_bussoft_ptr->bs_legacy_mem); + pci_add_resource_offset(&resources, &res[0], + prom_bussoft_ptr->bs_legacy_io); + pci_add_resource_offset(&resources, &res[1], + prom_bussoft_ptr->bs_legacy_mem); + + bus = pci_scan_root_bus(NULL, busnum, &pci_root_ops, controller, + &resources); + if (bus == NULL) { + kfree(res); + kfree(controller); } - return; - -error_return: - - kfree(sn_controller); - return; } -void sn_bus_store_sysdata(struct pci_dev *dev) +/* + * sn_bus_fixup + */ +void +sn_bus_fixup(struct pci_bus *bus) { - struct sysdata_el *element; + struct pci_dev *pci_dev = NULL; + struct pcibus_bussoft *prom_bussoft_ptr; - element = kzalloc(sizeof(struct sysdata_el), GFP_KERNEL); - if (!element) { - dev_dbg(dev, "%s: out of memory!\n", __FUNCTION__); - return; - } - element->sysdata = SN_PCIDEV_INFO(dev); - list_add(&element->entry, &sn_sysdata_list); -} + if (!bus->parent) { /* If root bus */ + prom_bussoft_ptr = PCI_CONTROLLER(bus)->platform_data; + if (prom_bussoft_ptr == NULL) { + printk(KERN_ERR + "sn_bus_fixup: 0x%04x:0x%02x Unable to " + "obtain prom_bussoft_ptr\n", + pci_domain_nr(bus), bus->number); + return; + } + sn_common_bus_fixup(bus, prom_bussoft_ptr); + } + list_for_each_entry(pci_dev, &bus->devices, bus_list) { + sn_io_slot_fixup(pci_dev); + } -void sn_bus_free_sysdata(void) -{ - struct sysdata_el *element; - struct list_head *list, *safe; - - list_for_each_safe(list, safe, &sn_sysdata_list) { - element = list_entry(list, struct sysdata_el, entry); - list_del(&element->entry); - list_del(&(((struct pcidev_info *) - (element->sysdata))->pdi_list)); - kfree(element->sysdata); - kfree(element); - } - return; } /* - * Ugly hack to get PCI setup until we have a proper ACPI namespace. + * sn_io_init - PROM does not have ACPI support to define nodes or root buses, + * so we need to do things the hard way, including initiating the + * bus scanning ourselves. */ -#define PCI_BUSES_TO_SCAN 256 - -static int __init sn_pci_init(void) +void __init sn_io_init(void) { int i, j; - struct pci_dev *pci_dev = NULL; - - if (!ia64_platform_is("sn2") || IS_RUNNING_ON_FAKE_PROM()) - return 0; - - /* - * prime sn_pci_provider[]. Individial provider init routines will - * override their respective default entries. - */ - - for (i = 0; i < PCIIO_ASIC_MAX_TYPES; i++) - sn_pci_provider[i] = &sn_pci_default_provider; - - pcibr_init_provider(); - tioca_init_provider(); - tioce_init_provider(); - /* - * This is needed to avoid bounce limit checks in the blk layer - */ - ia64_max_iommu_merge_mask = ~PAGE_MASK; sn_fixup_ionodes(); - sn_irq_lh_init(); - INIT_LIST_HEAD(&sn_sysdata_list); - sn_init_cpei_timer(); - -#ifdef CONFIG_PROC_FS - register_sn_procfs(); -#endif /* busses are not known yet ... */ for (i = 0; i <= max_segment_number; i++) for (j = 0; j <= max_pcibus_number; j++) sn_pci_controller_fixup(i, j, NULL); - - /* - * Generic Linux PCI Layer has created the pci_bus and pci_dev - * structures - time for us to add our SN PLatform specific - * information. - */ - - while ((pci_dev = - pci_get_device(PCI_ANY_ID, PCI_ANY_ID, pci_dev)) != NULL) - sn_pci_fixup_slot(pci_dev); - - sn_ioif_inited = 1; /* sn I/O infrastructure now initialized */ - - return 0; } - -/* - * hubdev_init_node() - Creates the HUB data structure and link them to it's - * own NODE specific data area. - */ -void hubdev_init_node(nodepda_t * npda, cnodeid_t node) -{ - struct hubdev_info *hubdev_info; - int size; - pg_data_t *pg; - - size = sizeof(struct hubdev_info); - - if (node >= num_online_nodes()) /* Headless/memless IO nodes */ - pg = NODE_DATA(0); - else - pg = NODE_DATA(node); - - hubdev_info = (struct hubdev_info *)alloc_bootmem_node(pg, size); - - npda->pdinfo = (void *)hubdev_info; -} - -geoid_t -cnodeid_get_geoid(cnodeid_t cnode) -{ - struct hubdev_info *hubdev; - - hubdev = (struct hubdev_info *)(NODEPDA(cnode)->pdinfo); - return hubdev->hdi_geoid; -} - -void sn_generate_path(struct pci_bus *pci_bus, char *address) -{ - nasid_t nasid; - cnodeid_t cnode; - geoid_t geoid; - moduleid_t moduleid; - u16 bricktype; - - nasid = NASID_GET(SN_PCIBUS_BUSSOFT(pci_bus)->bs_base); - cnode = nasid_to_cnodeid(nasid); - geoid = cnodeid_get_geoid(cnode); - moduleid = geo_module(geoid); - - sprintf(address, "module_%c%c%c%c%.2d", - '0'+RACK_GET_CLASS(MODULE_GET_RACK(moduleid)), - '0'+RACK_GET_GROUP(MODULE_GET_RACK(moduleid)), - '0'+RACK_GET_NUM(MODULE_GET_RACK(moduleid)), - MODULE_GET_BTCHAR(moduleid), MODULE_GET_BPOS(moduleid)); - - /* Tollhouse requires slot id to be displayed */ - bricktype = MODULE_GET_BTYPE(moduleid); - if ((bricktype == L1_BRICKTYPE_191010) || - (bricktype == L1_BRICKTYPE_1932)) - sprintf(address, "%s^%d", address, geo_slot(geoid)); -} - -subsys_initcall(sn_pci_init); -EXPORT_SYMBOL(sn_pci_fixup_slot); -EXPORT_SYMBOL(sn_pci_unfixup_slot); -EXPORT_SYMBOL(sn_pci_controller_fixup); -EXPORT_SYMBOL(sn_bus_store_sysdata); -EXPORT_SYMBOL(sn_bus_free_sysdata); -EXPORT_SYMBOL(sn_generate_path); diff --git a/arch/ia64/sn/kernel/iomv.c b/arch/ia64/sn/kernel/iomv.c index 7ce3cdad627..c77ebdf9811 100644 --- a/arch/ia64/sn/kernel/iomv.c +++ b/arch/ia64/sn/kernel/iomv.c @@ -1,12 +1,13 @@ -/* +/* * This file is subject to the terms and conditions of the GNU General Public * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2000-2003 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 2000-2003, 2006 Silicon Graphics, Inc. All rights reserved. */ #include <linux/module.h> +#include <linux/acpi.h> #include <asm/io.h> #include <asm/delay.h> #include <asm/vga.h> @@ -15,6 +16,7 @@ #include <asm/sn/pda.h> #include <asm/sn/sn_cpuid.h> #include <asm/sn/shub_mmr.h> +#include <asm/sn/acpi.h> #define IS_LEGACY_VGA_IOPORT(p) \ (((p) >= 0x3b0 && (p) <= 0x3bb) || ((p) >= 0x3c0 && (p) <= 0x3df)) @@ -24,18 +26,22 @@ * @port: port to convert * * Legacy in/out instructions are converted to ld/st instructions - * on IA64. This routine will convert a port number into a valid + * on IA64. This routine will convert a port number into a valid * SN i/o address. Used by sn_in*() and sn_out*(). */ + void *sn_io_addr(unsigned long port) { if (!IS_RUNNING_ON_SIMULATOR()) { if (IS_LEGACY_VGA_IOPORT(port)) - port += vga_console_iobase; + return (__ia64_mk_io_addr(port)); /* On sn2, legacy I/O ports don't point at anything */ if (port < (64 * 1024)) return NULL; - return ((void *)(port | __IA64_UNCACHED_OFFSET)); + if (SN_ACPI_BASE_SUPPORT()) + return (__ia64_mk_io_addr(port)); + else + return ((void *)(port | __IA64_UNCACHED_OFFSET)); } else { /* but the simulator uses them... */ unsigned long addr; @@ -57,7 +63,7 @@ EXPORT_SYMBOL(sn_io_addr); /** * __sn_mmiowb - I/O space memory barrier * - * See include/asm-ia64/io.h and Documentation/DocBook/deviceiobook.tmpl + * See arch/ia64/include/asm/io.h and Documentation/DocBook/deviceiobook.tmpl * for details. * * On SN2, we wait for the PIO_WRITE_STATUS SHub register to clear. diff --git a/arch/ia64/sn/kernel/irq.c b/arch/ia64/sn/kernel/irq.c index 7bb6ad188ba..85d09515490 100644 --- a/arch/ia64/sn/kernel/irq.c +++ b/arch/ia64/sn/kernel/irq.c @@ -5,12 +5,14 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (c) 2000-2006 Silicon Graphics, Inc. All Rights Reserved. + * Copyright (c) 2000-2008 Silicon Graphics, Inc. All Rights Reserved. */ #include <linux/irq.h> #include <linux/spinlock.h> #include <linux/init.h> +#include <linux/rculist.h> +#include <linux/slab.h> #include <asm/sn/addrs.h> #include <asm/sn/arch.h> #include <asm/sn/intr.h> @@ -19,12 +21,11 @@ #include <asm/sn/pcidev.h> #include <asm/sn/shub_mmr.h> #include <asm/sn/sn_sal.h> +#include <asm/sn/sn_feature_sets.h> -static void force_interrupt(int irq); static void register_intr_pda(struct sn_irq_info *sn_irq_info); static void unregister_intr_pda(struct sn_irq_info *sn_irq_info); -int sn_force_interrupt_flag = 1; extern int sn_ioif_inited; struct list_head **sn_irq_lh; static DEFINE_SPINLOCK(sn_irq_info_lock); /* non-IRQ lock */ @@ -59,80 +60,74 @@ void sn_intr_free(nasid_t local_nasid, int local_widget, (u64) sn_irq_info->irq_cookie, 0, 0); } -static unsigned int sn_startup_irq(unsigned int irq) +u64 sn_intr_redirect(nasid_t local_nasid, int local_widget, + struct sn_irq_info *sn_irq_info, + nasid_t req_nasid, int req_slice) +{ + struct ia64_sal_retval ret_stuff; + ret_stuff.status = 0; + ret_stuff.v0 = 0; + + SAL_CALL_NOLOCK(ret_stuff, (u64) SN_SAL_IOIF_INTERRUPT, + (u64) SAL_INTR_REDIRECT, (u64) local_nasid, + (u64) local_widget, __pa(sn_irq_info), + (u64) req_nasid, (u64) req_slice, 0); + + return ret_stuff.status; +} + +static unsigned int sn_startup_irq(struct irq_data *data) { return 0; } -static void sn_shutdown_irq(unsigned int irq) +static void sn_shutdown_irq(struct irq_data *data) { } -static void sn_disable_irq(unsigned int irq) +extern void ia64_mca_register_cpev(int); + +static void sn_disable_irq(struct irq_data *data) { + if (data->irq == local_vector_to_irq(IA64_CPE_VECTOR)) + ia64_mca_register_cpev(0); } -static void sn_enable_irq(unsigned int irq) +static void sn_enable_irq(struct irq_data *data) { + if (data->irq == local_vector_to_irq(IA64_CPE_VECTOR)) + ia64_mca_register_cpev(data->irq); } -static void sn_ack_irq(unsigned int irq) +static void sn_ack_irq(struct irq_data *data) { u64 event_occurred, mask; + unsigned int irq = data->irq & 0xff; - irq = irq & 0xff; event_occurred = HUB_L((u64*)LOCAL_MMR_ADDR(SH_EVENT_OCCURRED)); mask = event_occurred & SH_ALL_INT_MASK; HUB_S((u64*)LOCAL_MMR_ADDR(SH_EVENT_OCCURRED_ALIAS), mask); __set_bit(irq, (volatile void *)pda->sn_in_service_ivecs); - move_native_irq(irq); + irq_move_irq(data); } -static void sn_end_irq(unsigned int irq) -{ - int ivec; - u64 event_occurred; - - ivec = irq & 0xff; - if (ivec == SGI_UART_VECTOR) { - event_occurred = HUB_L((u64*)LOCAL_MMR_ADDR (SH_EVENT_OCCURRED)); - /* If the UART bit is set here, we may have received an - * interrupt from the UART that the driver missed. To - * make sure, we IPI ourselves to force us to look again. - */ - if (event_occurred & SH_EVENT_OCCURRED_UART_INT_MASK) { - platform_send_ipi(smp_processor_id(), SGI_UART_VECTOR, - IA64_IPI_DM_INT, 0); - } - } - __clear_bit(ivec, (volatile void *)pda->sn_in_service_ivecs); - if (sn_force_interrupt_flag) - force_interrupt(irq); -} - -static void sn_irq_info_free(struct rcu_head *head); - struct sn_irq_info *sn_retarget_vector(struct sn_irq_info *sn_irq_info, nasid_t nasid, int slice) { int vector; + int cpuid; +#ifdef CONFIG_SMP int cpuphys; +#endif int64_t bridge; int local_widget, status; nasid_t local_nasid; struct sn_irq_info *new_irq_info; struct sn_pcibus_provider *pci_provider; - new_irq_info = kmalloc(sizeof(struct sn_irq_info), GFP_ATOMIC); - if (new_irq_info == NULL) - return NULL; - - memcpy(new_irq_info, sn_irq_info, sizeof(struct sn_irq_info)); - - bridge = (u64) new_irq_info->irq_bridge; + bridge = (u64) sn_irq_info->irq_bridge; if (!bridge) { - kfree(new_irq_info); return NULL; /* irq is not a device interrupt */ } @@ -142,11 +137,26 @@ struct sn_irq_info *sn_retarget_vector(struct sn_irq_info *sn_irq_info, local_widget = TIO_SWIN_WIDGETNUM(bridge); else local_widget = SWIN_WIDGETNUM(bridge); - vector = sn_irq_info->irq_irq; + + /* Make use of SAL_INTR_REDIRECT if PROM supports it */ + status = sn_intr_redirect(local_nasid, local_widget, sn_irq_info, nasid, slice); + if (!status) { + new_irq_info = sn_irq_info; + goto finish_up; + } + + /* + * PROM does not support SAL_INTR_REDIRECT, or it failed. + * Revert to old method. + */ + new_irq_info = kmemdup(sn_irq_info, sizeof(struct sn_irq_info), + GFP_ATOMIC); + if (new_irq_info == NULL) + return NULL; + /* Free the old PROM new_irq_info structure */ sn_intr_free(local_nasid, local_widget, new_irq_info); - /* Update kernels new_irq_info with new target info */ unregister_intr_pda(new_irq_info); /* allocate a new PROM new_irq_info struct */ @@ -160,9 +170,18 @@ struct sn_irq_info *sn_retarget_vector(struct sn_irq_info *sn_irq_info, return NULL; } - cpuphys = nasid_slice_to_cpuid(nasid, slice); - new_irq_info->irq_cpuid = cpuphys; register_intr_pda(new_irq_info); + spin_lock(&sn_irq_info_lock); + list_replace_rcu(&sn_irq_info->list, &new_irq_info->list); + spin_unlock(&sn_irq_info_lock); + kfree_rcu(sn_irq_info, rcu); + + +finish_up: + /* Update kernels new_irq_info with new target info */ + cpuid = nasid_slice_to_cpuid(new_irq_info->irq_nasid, + new_irq_info->irq_slice); + new_irq_info->irq_cpuid = cpuid; pci_provider = sn_pci_provider[new_irq_info->irq_bridge_type]; @@ -174,43 +193,75 @@ struct sn_irq_info *sn_retarget_vector(struct sn_irq_info *sn_irq_info, pci_provider && pci_provider->target_interrupt) (pci_provider->target_interrupt)(new_irq_info); - spin_lock(&sn_irq_info_lock); - list_replace_rcu(&sn_irq_info->list, &new_irq_info->list); - spin_unlock(&sn_irq_info_lock); - call_rcu(&sn_irq_info->rcu, sn_irq_info_free); - #ifdef CONFIG_SMP + cpuphys = cpu_physical_id(cpuid); set_irq_affinity_info((vector & 0xff), cpuphys, 0); #endif return new_irq_info; } -static void sn_set_affinity_irq(unsigned int irq, cpumask_t mask) +static int sn_set_affinity_irq(struct irq_data *data, + const struct cpumask *mask, bool force) { struct sn_irq_info *sn_irq_info, *sn_irq_info_safe; + unsigned int irq = data->irq; nasid_t nasid; int slice; - nasid = cpuid_to_nasid(first_cpu(mask)); - slice = cpuid_to_slice(first_cpu(mask)); + nasid = cpuid_to_nasid(cpumask_first_and(mask, cpu_online_mask)); + slice = cpuid_to_slice(cpumask_first_and(mask, cpu_online_mask)); list_for_each_entry_safe(sn_irq_info, sn_irq_info_safe, sn_irq_lh[irq], list) (void)sn_retarget_vector(sn_irq_info, nasid, slice); + + return 0; } -struct hw_interrupt_type irq_type_sn = { - .typename = "SN hub", - .startup = sn_startup_irq, - .shutdown = sn_shutdown_irq, - .enable = sn_enable_irq, - .disable = sn_disable_irq, - .ack = sn_ack_irq, - .end = sn_end_irq, - .set_affinity = sn_set_affinity_irq +#ifdef CONFIG_SMP +void sn_set_err_irq_affinity(unsigned int irq) +{ + /* + * On systems which support CPU disabling (SHub2), all error interrupts + * are targeted at the boot CPU. + */ + if (is_shub2() && sn_prom_feature_available(PRF_CPU_DISABLE_SUPPORT)) + set_irq_affinity_info(irq, cpu_physical_id(0), 0); +} +#else +void sn_set_err_irq_affinity(unsigned int irq) { } +#endif + +static void +sn_mask_irq(struct irq_data *data) +{ +} + +static void +sn_unmask_irq(struct irq_data *data) +{ +} + +struct irq_chip irq_type_sn = { + .name = "SN hub", + .irq_startup = sn_startup_irq, + .irq_shutdown = sn_shutdown_irq, + .irq_enable = sn_enable_irq, + .irq_disable = sn_disable_irq, + .irq_ack = sn_ack_irq, + .irq_mask = sn_mask_irq, + .irq_unmask = sn_unmask_irq, + .irq_set_affinity = sn_set_affinity_irq }; +ia64_vector sn_irq_to_vector(int irq) +{ + if (irq >= IA64_NUM_VECTORS) + return 0; + return (ia64_vector)irq; +} + unsigned int sn_local_vector_to_irq(u8 vector) { return (CPU_VECTOR_TO_IRQ(smp_processor_id(), vector)); @@ -219,15 +270,13 @@ unsigned int sn_local_vector_to_irq(u8 vector) void sn_irq_init(void) { int i; - irq_desc_t *base_desc = irq_desc; ia64_first_device_vector = IA64_SN2_FIRST_DEVICE_VECTOR; ia64_last_device_vector = IA64_SN2_LAST_DEVICE_VECTOR; for (i = 0; i < NR_IRQS; i++) { - if (base_desc[i].chip == &no_irq_type) { - base_desc[i].chip = &irq_type_sn; - } + if (irq_get_chip(i) == &no_irq_chip) + irq_set_chip(i, &irq_type_sn); } } @@ -286,19 +335,14 @@ static void unregister_intr_pda(struct sn_irq_info *sn_irq_info) rcu_read_unlock(); } -static void sn_irq_info_free(struct rcu_head *head) -{ - struct sn_irq_info *sn_irq_info; - - sn_irq_info = container_of(head, struct sn_irq_info, rcu); - kfree(sn_irq_info); -} - void sn_irq_fixup(struct pci_dev *pci_dev, struct sn_irq_info *sn_irq_info) { nasid_t nasid = sn_irq_info->irq_nasid; int slice = sn_irq_info->irq_slice; int cpu = nasid_slice_to_cpuid(nasid, slice); +#ifdef CONFIG_SMP + int cpuphys; +#endif pci_dev_get(pci_dev); sn_irq_info->irq_cpuid = cpu; @@ -308,9 +352,20 @@ void sn_irq_fixup(struct pci_dev *pci_dev, struct sn_irq_info *sn_irq_info) spin_lock(&sn_irq_info_lock); list_add_rcu(&sn_irq_info->list, sn_irq_lh[sn_irq_info->irq_irq]); reserve_irq_vector(sn_irq_info->irq_irq); + if (sn_irq_info->irq_int_bit != -1) + irq_set_handler(sn_irq_info->irq_irq, handle_level_irq); spin_unlock(&sn_irq_info_lock); register_intr_pda(sn_irq_info); +#ifdef CONFIG_SMP + cpuphys = cpu_physical_id(cpu); + set_irq_affinity_info(sn_irq_info->irq_irq, cpuphys, 0); + /* + * Affinity was set by the PROM, prevent it from + * being reset by the request_irq() path. + */ + irqd_mark_affinity_was_set(irq_get_irq_data(sn_irq_info->irq_irq)); +#endif } void sn_irq_unfixup(struct pci_dev *pci_dev) @@ -335,7 +390,7 @@ void sn_irq_unfixup(struct pci_dev *pci_dev) spin_unlock(&sn_irq_info_lock); if (list_empty(sn_irq_lh[sn_irq_info->irq_irq])) free_irq_vector(sn_irq_info->irq_irq); - call_rcu(&sn_irq_info->rcu, sn_irq_info_free); + kfree_rcu(sn_irq_info, rcu); pci_dev_put(pci_dev); } @@ -346,22 +401,11 @@ sn_call_force_intr_provider(struct sn_irq_info *sn_irq_info) struct sn_pcibus_provider *pci_provider; pci_provider = sn_pci_provider[sn_irq_info->irq_bridge_type]; - if (pci_provider && pci_provider->force_interrupt) - (*pci_provider->force_interrupt)(sn_irq_info); -} - -static void force_interrupt(int irq) -{ - struct sn_irq_info *sn_irq_info; - - if (!sn_ioif_inited) - return; - - rcu_read_lock(); - list_for_each_entry_rcu(sn_irq_info, sn_irq_lh[irq], list) - sn_call_force_intr_provider(sn_irq_info); - rcu_read_unlock(); + /* Don't force an interrupt if the irq has been disabled */ + if (!irqd_irq_disabled(irq_get_irq_data(sn_irq_info->irq_irq)) && + pci_provider && pci_provider->force_interrupt) + (*pci_provider->force_interrupt)(sn_irq_info); } /* @@ -382,7 +426,7 @@ static void sn_check_intr(int irq, struct sn_irq_info *sn_irq_info) /* * Bridge types attached to TIO (anything but PIC) do not need this WAR * since they do not target Shub II interrupt registers. If that - * ever changes, this check needs to accomodate. + * ever changes, this check needs to accommodate. */ if (sn_irq_info->irq_bridge_type != PCIIO_ASIC_TYPE_PIC) return; diff --git a/arch/ia64/sn/kernel/mca.c b/arch/ia64/sn/kernel/mca.c index 3db62f24596..27793f7aa99 100644 --- a/arch/ia64/sn/kernel/mca.c +++ b/arch/ia64/sn/kernel/mca.c @@ -98,8 +98,9 @@ sn_platform_plat_specific_err_print(const u8 * sect_header, u8 ** oemdata, while (*sn_oemdata_size > sn_oemdata_bufsize) { u8 *newbuf = vmalloc(*sn_oemdata_size); if (!newbuf) { + mutex_unlock(&sn_oemdata_mutex); printk(KERN_ERR "%s: unable to extend sn_oemdata\n", - __FUNCTION__); + __func__); return 1; } vfree(*sn_oemdata); diff --git a/arch/ia64/sn/kernel/msi_sn.c b/arch/ia64/sn/kernel/msi_sn.c new file mode 100644 index 00000000000..afc58d2799a --- /dev/null +++ b/arch/ia64/sn/kernel/msi_sn.c @@ -0,0 +1,238 @@ +/* + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2006 Silicon Graphics, Inc. All Rights Reserved. + */ + +#include <linux/types.h> +#include <linux/irq.h> +#include <linux/pci.h> +#include <linux/cpumask.h> +#include <linux/msi.h> +#include <linux/slab.h> + +#include <asm/sn/addrs.h> +#include <asm/sn/intr.h> +#include <asm/sn/pcibus_provider_defs.h> +#include <asm/sn/pcidev.h> +#include <asm/sn/nodepda.h> + +struct sn_msi_info { + u64 pci_addr; + struct sn_irq_info *sn_irq_info; +}; + +static struct sn_msi_info sn_msi_info[NR_IRQS]; + +static struct irq_chip sn_msi_chip; + +void sn_teardown_msi_irq(unsigned int irq) +{ + nasid_t nasid; + int widget; + struct pci_dev *pdev; + struct pcidev_info *sn_pdev; + struct sn_irq_info *sn_irq_info; + struct pcibus_bussoft *bussoft; + struct sn_pcibus_provider *provider; + + sn_irq_info = sn_msi_info[irq].sn_irq_info; + if (sn_irq_info == NULL || sn_irq_info->irq_int_bit >= 0) + return; + + sn_pdev = (struct pcidev_info *)sn_irq_info->irq_pciioinfo; + pdev = sn_pdev->pdi_linux_pcidev; + provider = SN_PCIDEV_BUSPROVIDER(pdev); + + (*provider->dma_unmap)(pdev, + sn_msi_info[irq].pci_addr, + PCI_DMA_FROMDEVICE); + sn_msi_info[irq].pci_addr = 0; + + bussoft = SN_PCIDEV_BUSSOFT(pdev); + nasid = NASID_GET(bussoft->bs_base); + widget = (nasid & 1) ? + TIO_SWIN_WIDGETNUM(bussoft->bs_base) : + SWIN_WIDGETNUM(bussoft->bs_base); + + sn_intr_free(nasid, widget, sn_irq_info); + sn_msi_info[irq].sn_irq_info = NULL; + + destroy_irq(irq); +} + +int sn_setup_msi_irq(struct pci_dev *pdev, struct msi_desc *entry) +{ + struct msi_msg msg; + int widget; + int status; + nasid_t nasid; + u64 bus_addr; + struct sn_irq_info *sn_irq_info; + struct pcibus_bussoft *bussoft = SN_PCIDEV_BUSSOFT(pdev); + struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev); + int irq; + + if (!entry->msi_attrib.is_64) + return -EINVAL; + + if (bussoft == NULL) + return -EINVAL; + + if (provider == NULL || provider->dma_map_consistent == NULL) + return -EINVAL; + + irq = create_irq(); + if (irq < 0) + return irq; + + /* + * Set up the vector plumbing. Let the prom (via sn_intr_alloc) + * decide which cpu to direct this msi at by default. + */ + + nasid = NASID_GET(bussoft->bs_base); + widget = (nasid & 1) ? + TIO_SWIN_WIDGETNUM(bussoft->bs_base) : + SWIN_WIDGETNUM(bussoft->bs_base); + + sn_irq_info = kzalloc(sizeof(struct sn_irq_info), GFP_KERNEL); + if (! sn_irq_info) { + destroy_irq(irq); + return -ENOMEM; + } + + status = sn_intr_alloc(nasid, widget, sn_irq_info, irq, -1, -1); + if (status) { + kfree(sn_irq_info); + destroy_irq(irq); + return -ENOMEM; + } + + sn_irq_info->irq_int_bit = -1; /* mark this as an MSI irq */ + sn_irq_fixup(pdev, sn_irq_info); + + /* Prom probably should fill these in, but doesn't ... */ + sn_irq_info->irq_bridge_type = bussoft->bs_asic_type; + sn_irq_info->irq_bridge = (void *)bussoft->bs_base; + + /* + * Map the xio address into bus space + */ + bus_addr = (*provider->dma_map_consistent)(pdev, + sn_irq_info->irq_xtalkaddr, + sizeof(sn_irq_info->irq_xtalkaddr), + SN_DMA_MSI|SN_DMA_ADDR_XIO); + if (! bus_addr) { + sn_intr_free(nasid, widget, sn_irq_info); + kfree(sn_irq_info); + destroy_irq(irq); + return -ENOMEM; + } + + sn_msi_info[irq].sn_irq_info = sn_irq_info; + sn_msi_info[irq].pci_addr = bus_addr; + + msg.address_hi = (u32)(bus_addr >> 32); + msg.address_lo = (u32)(bus_addr & 0x00000000ffffffff); + + /* + * In the SN platform, bit 16 is a "send vector" bit which + * must be present in order to move the vector through the system. + */ + msg.data = 0x100 + irq; + + irq_set_msi_desc(irq, entry); + write_msi_msg(irq, &msg); + irq_set_chip_and_handler(irq, &sn_msi_chip, handle_edge_irq); + + return 0; +} + +#ifdef CONFIG_SMP +static int sn_set_msi_irq_affinity(struct irq_data *data, + const struct cpumask *cpu_mask, bool force) +{ + struct msi_msg msg; + int slice; + nasid_t nasid; + u64 bus_addr; + struct pci_dev *pdev; + struct pcidev_info *sn_pdev; + struct sn_irq_info *sn_irq_info; + struct sn_irq_info *new_irq_info; + struct sn_pcibus_provider *provider; + unsigned int cpu, irq = data->irq; + + cpu = cpumask_first_and(cpu_mask, cpu_online_mask); + sn_irq_info = sn_msi_info[irq].sn_irq_info; + if (sn_irq_info == NULL || sn_irq_info->irq_int_bit >= 0) + return -1; + + /* + * Release XIO resources for the old MSI PCI address + */ + + get_cached_msi_msg(irq, &msg); + sn_pdev = (struct pcidev_info *)sn_irq_info->irq_pciioinfo; + pdev = sn_pdev->pdi_linux_pcidev; + provider = SN_PCIDEV_BUSPROVIDER(pdev); + + bus_addr = (u64)(msg.address_hi) << 32 | (u64)(msg.address_lo); + (*provider->dma_unmap)(pdev, bus_addr, PCI_DMA_FROMDEVICE); + sn_msi_info[irq].pci_addr = 0; + + nasid = cpuid_to_nasid(cpu); + slice = cpuid_to_slice(cpu); + + new_irq_info = sn_retarget_vector(sn_irq_info, nasid, slice); + sn_msi_info[irq].sn_irq_info = new_irq_info; + if (new_irq_info == NULL) + return -1; + + /* + * Map the xio address into bus space + */ + + bus_addr = (*provider->dma_map_consistent)(pdev, + new_irq_info->irq_xtalkaddr, + sizeof(new_irq_info->irq_xtalkaddr), + SN_DMA_MSI|SN_DMA_ADDR_XIO); + + sn_msi_info[irq].pci_addr = bus_addr; + msg.address_hi = (u32)(bus_addr >> 32); + msg.address_lo = (u32)(bus_addr & 0x00000000ffffffff); + + write_msi_msg(irq, &msg); + cpumask_copy(data->affinity, cpu_mask); + + return 0; +} +#endif /* CONFIG_SMP */ + +static void sn_ack_msi_irq(struct irq_data *data) +{ + irq_move_irq(data); + ia64_eoi(); +} + +static int sn_msi_retrigger_irq(struct irq_data *data) +{ + unsigned int vector = data->irq; + ia64_resend_irq(vector); + + return 1; +} + +static struct irq_chip sn_msi_chip = { + .name = "PCI-MSI", + .irq_mask = mask_msi_irq, + .irq_unmask = unmask_msi_irq, + .irq_ack = sn_ack_msi_irq, +#ifdef CONFIG_SMP + .irq_set_affinity = sn_set_msi_irq_affinity, +#endif + .irq_retrigger = sn_msi_retrigger_irq, +}; diff --git a/arch/ia64/sn/kernel/setup.c b/arch/ia64/sn/kernel/setup.c index 5f2dcba7fa8..53b01b8e2f1 100644 --- a/arch/ia64/sn/kernel/setup.c +++ b/arch/ia64/sn/kernel/setup.c @@ -25,7 +25,6 @@ #include <linux/interrupt.h> #include <linux/acpi.h> #include <linux/compiler.h> -#include <linux/sched.h> #include <linux/root_dev.h> #include <linux/nodemask.h> #include <linux/pm.h> @@ -34,9 +33,9 @@ #include <asm/io.h> #include <asm/sal.h> #include <asm/machvec.h> -#include <asm/system.h> #include <asm/processor.h> #include <asm/vga.h> +#include <asm/setup.h> #include <asm/sn/arch.h> #include <asm/sn/addrs.h> #include <asm/sn/pda.h> @@ -65,8 +64,6 @@ extern void sn_timer_init(void); extern unsigned long last_time_offset; extern void (*ia64_mark_idle) (int); extern void snidle(int); -extern unsigned char acpi_kbd_controller_present; -extern unsigned long long (*ia64_printk_clock)(void); unsigned long sn_rtc_cycles_per_second; EXPORT_SYMBOL(sn_rtc_cycles_per_second); @@ -168,7 +165,7 @@ void __init early_sn_setup(void) * IO on SN2 is done via SAL calls, early_printk won't work without this. * * This code duplicates some of the ACPI table parsing that is in efi.c & sal.c. - * Any changes to those file may have to be made hereas well. + * Any changes to those file may have to be made here as well. */ efi_systab = (efi_system_table_t *) __va(ia64_boot_param->efi_systab); config_tables = __va(efi_systab->tables); @@ -195,7 +192,7 @@ void __init early_sn_setup(void) } extern int platform_intr_list[]; -static int __initdata shub_1_1_found; +static int shub_1_1_found; /* * sn_check_for_wars @@ -203,7 +200,7 @@ static int __initdata shub_1_1_found; * Set flag for enabling shub specific wars */ -static inline int __init is_shub_1_1(int nasid) +static inline int is_shub_1_1(int nasid) { unsigned long id; int rev; @@ -215,7 +212,7 @@ static inline int __init is_shub_1_1(int nasid) return rev <= 2; } -static void __init sn_check_for_wars(void) +static void sn_check_for_wars(void) { int cnode; @@ -244,7 +241,7 @@ static void __init sn_check_for_wars(void) * Note: This stuff is duped here because Altix requires the PCDP to * locate a usable VGA device due to lack of proper ACPI support. Structures * could be used from drivers/firmware/pcdp.h, but it was decided that moving - * this file to a more public location just for Altix use was undesireable. + * this file to a more public location just for Altix use was undesirable. */ struct hcdp_uart_desc { @@ -349,8 +346,7 @@ sn_scan_pcdp(void) continue; /* not PCI interconnect */ if (if_pci.translation & PCDP_PCI_TRANS_IOPORT) - vga_console_iobase = - if_pci.ioport_tra | __IA64_UNCACHED_OFFSET; + vga_console_iobase = if_pci.ioport_tra; if (if_pci.translation & PCDP_PCI_TRANS_MMIO) vga_console_membase = @@ -363,14 +359,6 @@ sn_scan_pcdp(void) static unsigned long sn2_rtc_initial; -static unsigned long long ia64_sn2_printk_clock(void) -{ - unsigned long rtc_now = rtc_time(); - - return (rtc_now - sn2_rtc_initial) * - (1000000000 / sn_rtc_cycles_per_second); -} - /** * sn_setup - SN platform setup routine * @cmdline_p: kernel command line @@ -389,7 +377,17 @@ void __init sn_setup(char **cmdline_p) ia64_sn_plat_set_error_handling_features(); // obsolete ia64_sn_set_os_feature(OSF_MCA_SLV_TO_OS_INIT_SLV); ia64_sn_set_os_feature(OSF_FEAT_LOG_SBES); + /* + * Note: The calls to notify the PROM of ACPI and PCI Segment + * support must be done prior to acpi_load_tables(), as + * an ACPI capable PROM will rebuild the DSDT as result + * of the call. + */ + ia64_sn_set_os_feature(OSF_PCISEGMENT_ENABLE); + ia64_sn_set_os_feature(OSF_ACPI_ENABLE); + /* Load the new DSDT and SSDT tables into the global table list. */ + acpi_table_init(); #if defined(CONFIG_VT) && defined(CONFIG_VGA_CONSOLE) /* @@ -414,6 +412,17 @@ void __init sn_setup(char **cmdline_p) if (! vga_console_membase) sn_scan_pcdp(); + /* + * Setup legacy IO space. + * vga_console_iobase maps to PCI IO Space address 0 on the + * bus containing the VGA console. + */ + if (vga_console_iobase) { + io_space[0].mmio_base = + (unsigned long) ioremap(vga_console_iobase, 0); + io_space[0].sparse = 0; + } + if (vga_console_membase) { /* usable vga ... make tty0 the preferred default console */ if (!strstr(*cmdline_p, "console=")) @@ -450,19 +459,6 @@ void __init sn_setup(char **cmdline_p) platform_intr_list[ACPI_INTERRUPT_CPEI] = IA64_CPE_VECTOR; - ia64_printk_clock = ia64_sn2_printk_clock; - - /* - * Old PROMs do not provide an ACPI FADT. Disable legacy keyboard - * support here so we don't have to listen to failed keyboard probe - * messages. - */ - if (is_shub1() && version <= 0x0209 && acpi_kbd_controller_present) { - printk(KERN_INFO "Disabling legacy keyboard support as prom " - "is too old and doesn't provide FADT\n"); - acpi_kbd_controller_present = 0; - } - printk("SGI SAL version %x.%02x\n", version >> 8, version & 0x00FF); /* @@ -511,12 +507,11 @@ static void __init sn_init_pdas(char **cmdline_p) cnodeid_t cnode; /* - * Allocate & initalize the nodepda for each node. + * Allocate & initialize the nodepda for each node. */ for_each_online_node(cnode) { nodepdaindr[cnode] = alloc_bootmem_node(NODE_DATA(cnode), sizeof(nodepda_t)); - memset(nodepdaindr[cnode], 0, sizeof(nodepda_t)); memset(nodepdaindr[cnode]->phys_cpuid, -1, sizeof(nodepdaindr[cnode]->phys_cpuid)); spin_lock_init(&nodepdaindr[cnode]->ptc_lock); @@ -525,11 +520,9 @@ static void __init sn_init_pdas(char **cmdline_p) /* * Allocate & initialize nodepda for TIOs. For now, put them on node 0. */ - for (cnode = num_online_nodes(); cnode < num_cnodes; cnode++) { + for (cnode = num_online_nodes(); cnode < num_cnodes; cnode++) nodepdaindr[cnode] = alloc_bootmem_node(NODE_DATA(0), sizeof(nodepda_t)); - memset(nodepdaindr[cnode], 0, sizeof(nodepda_t)); - } /* * Now copy the array of nodepda pointers to each nodepda. @@ -565,7 +558,7 @@ static void __init sn_init_pdas(char **cmdline_p) * Also sets up a few fields in the nodepda. Also known as * platform_cpu_init() by the ia64 machvec code. */ -void __cpuinit sn_cpu_init(void) +void sn_cpu_init(void) { int cpuid; int cpuphyid; @@ -574,7 +567,7 @@ void __cpuinit sn_cpu_init(void) int slice; int cnode; int i; - static int wars_have_been_checked; + static int wars_have_been_checked, set_cpu0_number; cpuid = smp_processor_id(); if (cpuid == 0 && IS_MEDUSA()) { @@ -599,8 +592,16 @@ void __cpuinit sn_cpu_init(void) /* * Don't check status. The SAL call is not supported on all PROMs * but a failure is harmless. + * Architecturally, cpu_init is always called twice on cpu 0. We + * should set cpu_number on cpu 0 once. */ - (void) ia64_sn_set_cpu_number(cpuid); + if (cpuid == 0) { + if (!set_cpu0_number) { + (void) ia64_sn_set_cpu_number(cpuid); + set_cpu0_number = 1; + } + } else + (void) ia64_sn_set_cpu_number(cpuid); /* * The boot cpu makes this call again after platform initialization is @@ -731,8 +732,7 @@ void __init build_cnode_tables(void) kl_config_hdr_t *klgraph_header; nasid = cnodeid_to_nasid(node); klgraph_header = ia64_sn_get_klconfig_addr(nasid); - if (klgraph_header == NULL) - BUG(); + BUG_ON(klgraph_header == NULL); brd = NODE_OFFSET_TO_LBOARD(nasid, klgraph_header->ch_board_info); while (brd) { if (board_needs_cnode(brd->brd_type) && physical_node_map[brd->brd_nasid] < 0) { @@ -749,7 +749,7 @@ nasid_slice_to_cpuid(int nasid, int slice) { long cpu; - for (cpu = 0; cpu < NR_CPUS; cpu++) + for (cpu = 0; cpu < nr_cpu_ids; cpu++) if (cpuid_to_nasid(cpu) == nasid && cpuid_to_slice(cpu) == slice) return cpu; @@ -763,5 +763,13 @@ int sn_prom_feature_available(int id) return 0; return test_bit(id, sn_prom_features); } + +void +sn_kernel_launch_event(void) +{ + /* ignore status until we understand possible failure, if any*/ + if (ia64_sn_kernel_launch_event()) + printk(KERN_ERR "KEXEC is not supported in this PROM, Please update the PROM.\n"); +} EXPORT_SYMBOL(sn_prom_feature_available); diff --git a/arch/ia64/sn/kernel/sn2/Makefile b/arch/ia64/sn/kernel/sn2/Makefile index 99e17769323..3d09108d427 100644 --- a/arch/ia64/sn/kernel/sn2/Makefile +++ b/arch/ia64/sn/kernel/sn2/Makefile @@ -9,7 +9,7 @@ # sn2 specific kernel files # -CPPFLAGS += -I$(srctree)/arch/ia64/sn/include +ccflags-y := -Iarch/ia64/sn/include obj-y += cache.o io.o ptc_deadlock.o sn2_smp.o sn_proc_fs.o \ prominfo_proc.o timer.o timer_interrupt.o sn_hwperf.o diff --git a/arch/ia64/sn/kernel/sn2/prominfo_proc.c b/arch/ia64/sn/kernel/sn2/prominfo_proc.c index 4dcce3d0e04..ec4de2b0965 100644 --- a/arch/ia64/sn/kernel/sn2/prominfo_proc.c +++ b/arch/ia64/sn/kernel/sn2/prominfo_proc.c @@ -11,8 +11,8 @@ #include <linux/module.h> #include <linux/slab.h> #include <linux/proc_fs.h> +#include <linux/seq_file.h> #include <linux/nodemask.h> -#include <asm/system.h> #include <asm/io.h> #include <asm/sn/sn_sal.h> #include <asm/sn/sn_cpuid.h> @@ -102,18 +102,18 @@ get_fit_entry(unsigned long nasid, int index, unsigned long *fentry, /* * These two routines display the FIT table for each node. */ -static int dump_fit_entry(char *page, unsigned long *fentry) +static void dump_fit_entry(struct seq_file *m, unsigned long *fentry) { unsigned type; type = FIT_TYPE(fentry[1]); - return sprintf(page, "%02x %-25s %x.%02x %016lx %u\n", - type, - fit_type_name(type), - FIT_MAJOR(fentry[1]), FIT_MINOR(fentry[1]), - fentry[0], - /* mult by sixteen to get size in bytes */ - (unsigned)(fentry[1] & 0xffffff) * 16); + seq_printf(m, "%02x %-25s %x.%02x %016lx %u\n", + type, + fit_type_name(type), + FIT_MAJOR(fentry[1]), FIT_MINOR(fentry[1]), + fentry[0], + /* mult by sixteen to get size in bytes */ + (unsigned)(fentry[1] & 0xffffff) * 16); } @@ -125,31 +125,39 @@ static int dump_fit_entry(char *page, unsigned long *fentry) * OK except for 4kB pages (and no one is going to do that on SN * anyway). */ -static int -dump_fit(char *page, unsigned long nasid) +static int proc_fit_show(struct seq_file *m, void *v) { + unsigned long nasid = (unsigned long)m->private; unsigned long fentry[2]; int index; - char *p; - p = page; for (index=0;;index++) { BUG_ON(index * 60 > PAGE_SIZE); if (get_fit_entry(nasid, index, fentry, NULL, 0)) break; - p += dump_fit_entry(p, fentry); + dump_fit_entry(m, fentry); } + return 0; +} - return p - page; +static int proc_fit_open(struct inode *inode, struct file *file) +{ + return single_open(file, proc_fit_show, PDE_DATA(inode)); } -static int -dump_version(char *page, unsigned long nasid) +static const struct file_operations proc_fit_fops = { + .open = proc_fit_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static int proc_version_show(struct seq_file *m, void *v) { + unsigned long nasid = (unsigned long)m->private; unsigned long fentry[2]; char banner[128]; int index; - int len; for (index = 0; ; index++) { if (get_fit_entry(nasid, index, fentry, banner, @@ -159,56 +167,24 @@ dump_version(char *page, unsigned long nasid) break; } - len = sprintf(page, "%x.%02x\n", FIT_MAJOR(fentry[1]), - FIT_MINOR(fentry[1])); - page += len; + seq_printf(m, "%x.%02x\n", FIT_MAJOR(fentry[1]), FIT_MINOR(fentry[1])); if (banner[0]) - len += snprintf(page, PAGE_SIZE-len, "%s\n", banner); - - return len; -} - -/* same as in proc_misc.c */ -static int -proc_calc_metrics(char *page, char **start, off_t off, int count, int *eof, - int len) -{ - if (len <= off + count) - *eof = 1; - *start = page + off; - len -= off; - if (len > count) - len = count; - if (len < 0) - len = 0; - return len; + seq_printf(m, "%s\n", banner); + return 0; } -static int -read_version_entry(char *page, char **start, off_t off, int count, int *eof, - void *data) +static int proc_version_open(struct inode *inode, struct file *file) { - int len; - - /* data holds the NASID of the node */ - len = dump_version(page, (unsigned long)data); - len = proc_calc_metrics(page, start, off, count, eof, len); - return len; + return single_open(file, proc_version_show, PDE_DATA(inode)); } -static int -read_fit_entry(char *page, char **start, off_t off, int count, int *eof, - void *data) -{ - int len; - - /* data holds the NASID of the node */ - len = dump_fit(page, (unsigned long)data); - len = proc_calc_metrics(page, start, off, count, eof, len); - - return len; -} +static const struct file_operations proc_version_fops = { + .open = proc_version_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; /* module entry points */ int __init prominfo_init(void); @@ -217,63 +193,39 @@ void __exit prominfo_exit(void); module_init(prominfo_init); module_exit(prominfo_exit); -static struct proc_dir_entry **proc_entries; -static struct proc_dir_entry *sgi_prominfo_entry; - #define NODE_NAME_LEN 11 int __init prominfo_init(void) { - struct proc_dir_entry **entp; - struct proc_dir_entry *p; + struct proc_dir_entry *sgi_prominfo_entry; cnodeid_t cnodeid; - unsigned long nasid; - int size; - char name[NODE_NAME_LEN]; if (!ia64_platform_is("sn2")) return 0; - size = num_online_nodes() * sizeof(struct proc_dir_entry *); - proc_entries = kzalloc(size, GFP_KERNEL); - if (!proc_entries) - return -ENOMEM; - sgi_prominfo_entry = proc_mkdir("sgi_prominfo", NULL); + if (!sgi_prominfo_entry) + return -ENOMEM; - entp = proc_entries; for_each_online_node(cnodeid) { + struct proc_dir_entry *dir; + unsigned long nasid; + char name[NODE_NAME_LEN]; + sprintf(name, "node%d", cnodeid); - *entp = proc_mkdir(name, sgi_prominfo_entry); + dir = proc_mkdir(name, sgi_prominfo_entry); + if (!dir) + continue; nasid = cnodeid_to_nasid(cnodeid); - p = create_proc_read_entry("fit", 0, *entp, read_fit_entry, - (void *)nasid); - if (p) - p->owner = THIS_MODULE; - p = create_proc_read_entry("version", 0, *entp, - read_version_entry, (void *)nasid); - if (p) - p->owner = THIS_MODULE; - entp++; + proc_create_data("fit", 0, dir, + &proc_fit_fops, (void *)nasid); + proc_create_data("version", 0, dir, + &proc_version_fops, (void *)nasid); } - return 0; } void __exit prominfo_exit(void) { - struct proc_dir_entry **entp; - unsigned int cnodeid; - char name[NODE_NAME_LEN]; - - entp = proc_entries; - for_each_online_node(cnodeid) { - remove_proc_entry("fit", *entp); - remove_proc_entry("version", *entp); - sprintf(name, "node%d", cnodeid); - remove_proc_entry(name, sgi_prominfo_entry); - entp++; - } - remove_proc_entry("sgi_prominfo", NULL); - kfree(proc_entries); + remove_proc_subtree("sgi_prominfo", NULL); } diff --git a/arch/ia64/sn/kernel/sn2/sn2_smp.c b/arch/ia64/sn/kernel/sn2/sn2_smp.c index d9d306c79f2..68c84541162 100644 --- a/arch/ia64/sn/kernel/sn2/sn2_smp.c +++ b/arch/ia64/sn/kernel/sn2/sn2_smp.c @@ -26,7 +26,6 @@ #include <asm/processor.h> #include <asm/irq.h> #include <asm/sal.h> -#include <asm/system.h> #include <asm/delay.h> #include <asm/io.h> #include <asm/smp.h> @@ -40,12 +39,16 @@ #include <asm/sn/shub_mmr.h> #include <asm/sn/nodepda.h> #include <asm/sn/rw_mmr.h> +#include <asm/sn/sn_feature_sets.h> DEFINE_PER_CPU(struct ptc_stats, ptcstats); DECLARE_PER_CPU(struct ptc_stats, ptcstats); static __cacheline_aligned DEFINE_SPINLOCK(sn2_global_ptc_lock); +/* 0 = old algorithm (no IPI flushes), 1 = ipi deadlock flush, 2 = ipi instead of SHUB ptc, >2 = always ipi */ +static int sn2_flush_opt = 0; + extern unsigned long sn2_ptc_deadlock_recovery_core(volatile unsigned long *, unsigned long, volatile unsigned long *, unsigned long, @@ -76,6 +79,8 @@ struct ptc_stats { unsigned long shub_itc_clocks; unsigned long shub_itc_clocks_max; unsigned long shub_ptc_flushes_not_my_mm; + unsigned long shub_ipi_flushes; + unsigned long shub_ipi_flushes_itc_clocks; }; #define sn2_ptctest 0 @@ -99,7 +104,7 @@ static inline unsigned long wait_piowc(void) * * SN2 PIO writes from separate CPUs are not guaranteed to arrive in order. * Context switching user threads which have memory-mapped MMIO may cause - * PIOs to issue from seperate CPUs, thus the PIO writes must be drained + * PIOs to issue from separate CPUs, thus the PIO writes must be drained * from the previous CPU's Shub before execution resumes on the new CPU. */ void sn_migrate(struct task_struct *task) @@ -121,6 +126,18 @@ void sn_tlb_migrate_finish(struct mm_struct *mm) flush_tlb_mm(mm); } +static void +sn2_ipi_flush_all_tlb(struct mm_struct *mm) +{ + unsigned long itc; + + itc = ia64_get_itc(); + smp_flush_tlb_cpumask(*mm_cpumask(mm)); + itc = ia64_get_itc() - itc; + __get_cpu_var(ptcstats).shub_ipi_flushes_itc_clocks += itc; + __get_cpu_var(ptcstats).shub_ipi_flushes++; +} + /** * sn2_global_tlb_purge - globally purge translation cache of virtual address range * @mm: mm_struct containing virtual address range @@ -154,12 +171,17 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, unsigned long itc, itc2, flags, data0 = 0, data1 = 0, rr_value, old_rr = 0; short nasids[MAX_NUMNODES], nix; nodemask_t nodes_flushed; - int active, max_active, deadlock; + int active, max_active, deadlock, flush_opt = sn2_flush_opt; + + if (flush_opt > 2) { + sn2_ipi_flush_all_tlb(mm); + return; + } nodes_clear(nodes_flushed); i = 0; - for_each_cpu_mask(cpu, mm->cpu_vm_mask) { + for_each_cpu(cpu, mm_cpumask(mm)) { cnode = cpu_to_node(cpu); node_set(cnode, nodes_flushed); lcpu = cpu; @@ -189,6 +211,12 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, return; } + if (flush_opt == 2) { + sn2_ipi_flush_all_tlb(mm); + preempt_enable(); + return; + } + itc = ia64_get_itc(); nix = 0; for_each_node_mask(cnode, nodes_flushed) @@ -256,6 +284,8 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, } if (active >= max_active || i == (nix - 1)) { if ((deadlock = wait_piowc())) { + if (flush_opt == 1) + goto done; sn2_ptc_deadlock_recovery(nasids, ibegin, i, mynasid, ptc0, data0, ptc1, data1); if (reset_max_active_on_deadlock()) max_active = 1; @@ -267,6 +297,7 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, start += (1UL << nbits); } while (start < end); +done: itc2 = ia64_get_itc() - itc2; __get_cpu_var(ptcstats).shub_itc_clocks += itc2; if (itc2 > __get_cpu_var(ptcstats).shub_itc_clocks_max) @@ -279,6 +310,11 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, spin_unlock_irqrestore(PTC_LOCK(shub1), flags); + if (flush_opt == 1 && deadlock) { + __get_cpu_var(ptcstats).deadlocks++; + sn2_ipi_flush_all_tlb(mm); + } + preempt_enable(); } @@ -393,13 +429,38 @@ void sn2_send_IPI(int cpuid, int vector, int delivery_mode, int redirect) sn_send_IPI_phys(nasid, physid, vector, delivery_mode); } +#ifdef CONFIG_HOTPLUG_CPU +/** + * sn_cpu_disable_allowed - Determine if a CPU can be disabled. + * @cpu - CPU that is requested to be disabled. + * + * CPU disable is only allowed on SHub2 systems running with a PROM + * that supports CPU disable. It is not permitted to disable the boot processor. + */ +bool sn_cpu_disable_allowed(int cpu) +{ + if (is_shub2() && sn_prom_feature_available(PRF_CPU_DISABLE_SUPPORT)) { + if (cpu != 0) + return true; + else + printk(KERN_WARNING + "Disabling the boot processor is not allowed.\n"); + + } else + printk(KERN_WARNING + "CPU disable is not supported on this system.\n"); + + return false; +} +#endif /* CONFIG_HOTPLUG_CPU */ + #ifdef CONFIG_PROC_FS #define PTC_BASENAME "sgi_sn/ptc_statistics" static void *sn2_ptc_seq_start(struct seq_file *file, loff_t * offset) { - if (*offset < NR_CPUS) + if (*offset < nr_cpu_ids) return offset; return NULL; } @@ -407,7 +468,7 @@ static void *sn2_ptc_seq_start(struct seq_file *file, loff_t * offset) static void *sn2_ptc_seq_next(struct seq_file *file, void *data, loff_t * offset) { (*offset)++; - if (*offset < NR_CPUS) + if (*offset < nr_cpu_ids) return offset; return NULL; } @@ -425,25 +486,45 @@ static int sn2_ptc_seq_show(struct seq_file *file, void *data) if (!cpu) { seq_printf(file, - "# cpu ptc_l newrid ptc_flushes nodes_flushed deadlocks lock_nsec shub_nsec shub_nsec_max not_my_mm deadlock2\n"); - seq_printf(file, "# ptctest %d\n", sn2_ptctest); + "# cpu ptc_l newrid ptc_flushes nodes_flushed deadlocks lock_nsec shub_nsec shub_nsec_max not_my_mm deadlock2 ipi_fluches ipi_nsec\n"); + seq_printf(file, "# ptctest %d, flushopt %d\n", sn2_ptctest, sn2_flush_opt); } - if (cpu < NR_CPUS && cpu_online(cpu)) { + if (cpu < nr_cpu_ids && cpu_online(cpu)) { stat = &per_cpu(ptcstats, cpu); - seq_printf(file, "cpu %d %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld\n", cpu, stat->ptc_l, + seq_printf(file, "cpu %d %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld\n", cpu, stat->ptc_l, stat->change_rid, stat->shub_ptc_flushes, stat->nodes_flushed, stat->deadlocks, - 1000 * stat->lock_itc_clocks / per_cpu(cpu_info, cpu).cyc_per_usec, - 1000 * stat->shub_itc_clocks / per_cpu(cpu_info, cpu).cyc_per_usec, - 1000 * stat->shub_itc_clocks_max / per_cpu(cpu_info, cpu).cyc_per_usec, + 1000 * stat->lock_itc_clocks / per_cpu(ia64_cpu_info, cpu).cyc_per_usec, + 1000 * stat->shub_itc_clocks / per_cpu(ia64_cpu_info, cpu).cyc_per_usec, + 1000 * stat->shub_itc_clocks_max / per_cpu(ia64_cpu_info, cpu).cyc_per_usec, stat->shub_ptc_flushes_not_my_mm, - stat->deadlocks2); + stat->deadlocks2, + stat->shub_ipi_flushes, + 1000 * stat->shub_ipi_flushes_itc_clocks / per_cpu(ia64_cpu_info, cpu).cyc_per_usec); } return 0; } -static struct seq_operations sn2_ptc_seq_ops = { +static ssize_t sn2_ptc_proc_write(struct file *file, const char __user *user, size_t count, loff_t *data) +{ + int cpu; + char optstr[64]; + + if (count == 0 || count > sizeof(optstr)) + return -EINVAL; + if (copy_from_user(optstr, user, count)) + return -EFAULT; + optstr[count - 1] = '\0'; + sn2_flush_opt = simple_strtoul(optstr, NULL, 0); + + for_each_online_cpu(cpu) + memset(&per_cpu(ptcstats, cpu), 0, sizeof(struct ptc_stats)); + + return count; +} + +static const struct seq_operations sn2_ptc_seq_ops = { .start = sn2_ptc_seq_start, .next = sn2_ptc_seq_next, .stop = sn2_ptc_seq_stop, @@ -455,9 +536,10 @@ static int sn2_ptc_proc_open(struct inode *inode, struct file *file) return seq_open(file, &sn2_ptc_seq_ops); } -static struct file_operations proc_sn2_ptc_operations = { +static const struct file_operations proc_sn2_ptc_operations = { .open = sn2_ptc_proc_open, .read = seq_read, + .write = sn2_ptc_proc_write, .llseek = seq_lseek, .release = seq_release, }; @@ -469,11 +551,12 @@ static int __init sn2_ptc_init(void) if (!ia64_platform_is("sn2")) return 0; - if (!(proc_sn2_ptc = create_proc_entry(PTC_BASENAME, 0444, NULL))) { + proc_sn2_ptc = proc_create(PTC_BASENAME, 0444, + NULL, &proc_sn2_ptc_operations); + if (!proc_sn2_ptc) { printk(KERN_ERR "unable to create %s proc entry", PTC_BASENAME); return -EINVAL; } - proc_sn2_ptc->proc_fops = &proc_sn2_ptc_operations; spin_lock_init(&sn2_global_ptc_lock); return 0; } diff --git a/arch/ia64/sn/kernel/sn2/sn_hwperf.c b/arch/ia64/sn/kernel/sn2/sn_hwperf.c index b632b9c1e3b..b9992571c03 100644 --- a/arch/ia64/sn/kernel/sn2/sn_hwperf.c +++ b/arch/ia64/sn/kernel/sn2/sn_hwperf.c @@ -25,18 +25,18 @@ #include <linux/fs.h> #include <linux/slab.h> +#include <linux/export.h> #include <linux/vmalloc.h> #include <linux/seq_file.h> #include <linux/miscdevice.h> #include <linux/utsname.h> #include <linux/cpumask.h> -#include <linux/smp_lock.h> #include <linux/nodemask.h> #include <linux/smp.h> +#include <linux/mutex.h> #include <asm/processor.h> #include <asm/topology.h> -#include <asm/semaphore.h> #include <asm/uaccess.h> #include <asm/sal.h> #include <asm/sn/io.h> @@ -50,7 +50,7 @@ static void *sn_hwperf_salheap = NULL; static int sn_hwperf_obj_cnt = 0; static nasid_t sn_hwperf_master_nasid = INVALID_NASID; static int sn_hwperf_init(void); -static DECLARE_MUTEX(sn_hwperf_init_mutex); +static DEFINE_MUTEX(sn_hwperf_init_mutex); #define cnode_possible(n) ((n) < num_cnodes) @@ -66,7 +66,8 @@ static int sn_hwperf_enum_objects(int *nobj, struct sn_hwperf_object_info **ret) } sz = sn_hwperf_obj_cnt * sizeof(struct sn_hwperf_object_info); - if ((objbuf = (struct sn_hwperf_object_info *) vmalloc(sz)) == NULL) { + objbuf = vmalloc(sz); + if (objbuf == NULL) { printk("sn_hwperf_enum_objects: vmalloc(%d) failed\n", (int)sz); e = -ENOMEM; goto out; @@ -189,7 +190,7 @@ static void print_pci_topology(struct seq_file *s) int e; for (sz = PAGE_SIZE; sz < 16 * PAGE_SIZE; sz += PAGE_SIZE) { - if (!(p = (char *)kmalloc(sz, GFP_KERNEL))) + if (!(p = kmalloc(sz, GFP_KERNEL))) break; e = ia64_sn_ioif_get_pci_topology(__pa(p), sz); if (e == SALRET_OK) @@ -274,8 +275,7 @@ static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objb /* get it's interconnect topology */ sz = op->ports * sizeof(struct sn_hwperf_port_info); - if (sz > sizeof(ptdata)) - BUG(); + BUG_ON(sz > sizeof(ptdata)); e = ia64_sn_hwperf_op(sn_hwperf_master_nasid, SN_HWPERF_ENUM_PORTS, nodeobj->id, sz, (u64)&ptdata, 0, 0, NULL); @@ -309,8 +309,7 @@ static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objb if (router && (!found_cpu || !found_mem)) { /* search for a node connected to the same router */ sz = router->ports * sizeof(struct sn_hwperf_port_info); - if (sz > sizeof(ptdata)) - BUG(); + BUG_ON(sz > sizeof(ptdata)); e = ia64_sn_hwperf_op(sn_hwperf_master_nasid, SN_HWPERF_ENUM_PORTS, router->id, sz, (u64)&ptdata, 0, 0, NULL); @@ -384,7 +383,6 @@ static int sn_topology_show(struct seq_file *s, void *d) int j; const char *slabname; int ordinal; - cpumask_t cpumask; char slice; struct cpuinfo_ia64 *c; struct sn_hwperf_port_info *ptdata; @@ -416,14 +414,14 @@ static int sn_topology_show(struct seq_file *s, void *d) } seq_printf(s, "partition %u %s local " "shubtype %s, " - "nasid_mask 0x%016lx, " + "nasid_mask 0x%016llx, " "nasid_bits %d:%d, " "system_size %d, " "sharing_size %d, " "coherency_domain %d, " "region_size %d\n", - partid, system_utsname.nodename, + partid, utsname()->nodename, shubtype ? "shub2" : "shub1", (u64)nasid_mask << nasid_shift, nasid_msb, nasid_shift, system_size, sharing_size, coher, region_size); @@ -472,23 +470,21 @@ static int sn_topology_show(struct seq_file *s, void *d) * CPUs on this node, if any */ if (!SN_HWPERF_IS_IONODE(obj)) { - cpumask = node_to_cpumask(ordinal); - for_each_online_cpu(i) { - if (cpu_isset(i, cpumask)) { - slice = 'a' + cpuid_to_slice(i); - c = cpu_data(i); - seq_printf(s, "cpu %d %s%c local" - " freq %luMHz, arch ia64", - i, obj->location, slice, - c->proc_freq / 1000000); - for_each_online_cpu(j) { - seq_printf(s, j ? ":%d" : ", dist %d", - node_distance( + for_each_cpu_and(i, cpu_online_mask, + cpumask_of_node(ordinal)) { + slice = 'a' + cpuid_to_slice(i); + c = cpu_data(i); + seq_printf(s, "cpu %d %s%c local" + " freq %luMHz, arch ia64", + i, obj->location, slice, + c->proc_freq / 1000000); + for_each_online_cpu(j) { + seq_printf(s, j ? ":%d" : ", dist %d", + node_distance( cpu_to_node(i), cpu_to_node(j))); - } - seq_putc(s, '\n'); } + seq_putc(s, '\n'); } } } @@ -576,7 +572,7 @@ static void sn_topology_stop(struct seq_file *m, void *v) /* * /proc/sgi_sn/sn_topology, read-only using seq_file */ -static struct seq_operations sn_topology_seq_ops = { +static const struct seq_operations sn_topology_seq_ops = { .start = sn_topology_start, .next = sn_topology_next, .stop = sn_topology_stop, @@ -614,28 +610,32 @@ static int sn_hwperf_op_cpu(struct sn_hwperf_op_info *op_info) op_info->a->arg &= SN_HWPERF_ARG_OBJID_MASK; if (cpu != SN_HWPERF_ARG_ANY_CPU) { - if (cpu >= NR_CPUS || !cpu_online(cpu)) { + if (cpu >= nr_cpu_ids || !cpu_online(cpu)) { r = -EINVAL; goto out; } } - if (cpu == SN_HWPERF_ARG_ANY_CPU || cpu == get_cpu()) { - /* don't care, or already on correct cpu */ + if (cpu == SN_HWPERF_ARG_ANY_CPU) { + /* don't care which cpu */ sn_hwperf_call_sal(op_info); - } - else { + } else if (cpu == get_cpu()) { + /* already on correct cpu */ + sn_hwperf_call_sal(op_info); + put_cpu(); + } else { + put_cpu(); if (use_ipi) { /* use an interprocessor interrupt to call SAL */ smp_call_function_single(cpu, sn_hwperf_call_sal, - op_info, 1, 1); + op_info, 1); } else { /* migrate the task before calling SAL */ save_allowed = current->cpus_allowed; - set_cpus_allowed(current, cpumask_of_cpu(cpu)); + set_cpus_allowed_ptr(current, cpumask_of(cpu)); sn_hwperf_call_sal(op_info); - set_cpus_allowed(current, save_allowed); + set_cpus_allowed_ptr(current, &save_allowed); } } r = op_info->ret; @@ -686,8 +686,7 @@ static int sn_hwperf_map_err(int hwperf_err) /* * ioctl for "sn_hwperf" misc device */ -static int -sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg) +static long sn_hwperf_ioctl(struct file *fp, u32 op, unsigned long arg) { struct sn_hwperf_ioctl_args a; struct cpuinfo_ia64 *cdata; @@ -703,8 +702,6 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg) int i; int j; - unlock_kernel(); - /* only user requests are allowed here */ if ((op & SN_HWPERF_OP_MASK) < 10) { r = -EINVAL; @@ -750,9 +747,10 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg) goto error; } else if ((r = sn_hwperf_enum_objects(&nobj, &objs)) == 0) { + int cpuobj_index = 0; + memset(p, 0, a.sz); for (i = 0; i < nobj; i++) { - int cpuobj_index = 0; if (!SN_HWPERF_IS_NODE(objs + i)) continue; node = sn_hwperf_obj_to_cnode(objs + i); @@ -789,17 +787,18 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg) break; case SN_HWPERF_GET_OBJ_NODE: - if (a.sz != sizeof(u64) || a.arg < 0) { + i = a.arg; + if (a.sz != sizeof(u64) || i < 0) { r = -EINVAL; goto error; } if ((r = sn_hwperf_enum_objects(&nobj, &objs)) == 0) { - if (a.arg >= nobj) { + if (i >= nobj) { r = -EINVAL; vfree(objs); goto error; } - if (objs[(i = a.arg)].id != a.arg) { + if (objs[i].id != a.arg) { for (i = 0; i < nobj; i++) { if (objs[i].id == a.arg) break; @@ -861,12 +860,12 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg) error: vfree(p); - lock_kernel(); return r; } -static struct file_operations sn_hwperf_fops = { - .ioctl = sn_hwperf_ioctl, +static const struct file_operations sn_hwperf_fops = { + .unlocked_ioctl = sn_hwperf_ioctl, + .llseek = noop_llseek, }; static struct miscdevice sn_hwperf_dev = { @@ -882,10 +881,10 @@ static int sn_hwperf_init(void) int e = 0; /* single threaded, once-only initialization */ - down(&sn_hwperf_init_mutex); + mutex_lock(&sn_hwperf_init_mutex); if (sn_hwperf_salheap) { - up(&sn_hwperf_init_mutex); + mutex_unlock(&sn_hwperf_init_mutex); return e; } @@ -934,7 +933,7 @@ out: sn_hwperf_salheap = NULL; sn_hwperf_obj_cnt = 0; } - up(&sn_hwperf_init_mutex); + mutex_unlock(&sn_hwperf_init_mutex); return e; } @@ -978,7 +977,7 @@ int sn_hwperf_get_nearest_node(cnodeid_t node, return e; } -static int __devinit sn_hwperf_misc_register_init(void) +static int sn_hwperf_misc_register_init(void) { int e; diff --git a/arch/ia64/sn/kernel/sn2/sn_proc_fs.c b/arch/ia64/sn/kernel/sn2/sn_proc_fs.c index 43ddc2eccb9..7aab87f4806 100644 --- a/arch/ia64/sn/kernel/sn2/sn_proc_fs.c +++ b/arch/ia64/sn/kernel/sn2/sn_proc_fs.c @@ -36,7 +36,7 @@ static int system_serial_number_open(struct inode *inode, struct file *file) static int licenseID_show(struct seq_file *s, void *p) { - seq_printf(s, "0x%lx\n", sn_partition_serial_number_val()); + seq_printf(s, "0x%llx\n", sn_partition_serial_number_val()); return 0; } @@ -45,38 +45,6 @@ static int licenseID_open(struct inode *inode, struct file *file) return single_open(file, licenseID_show, NULL); } -/* - * Enable forced interrupt by default. - * When set, the sn interrupt handler writes the force interrupt register on - * the bridge chip. The hardware will then send an interrupt message if the - * interrupt line is active. This mimics a level sensitive interrupt. - */ -extern int sn_force_interrupt_flag; - -static int sn_force_interrupt_show(struct seq_file *s, void *p) -{ - seq_printf(s, "Force interrupt is %s\n", - sn_force_interrupt_flag ? "enabled" : "disabled"); - return 0; -} - -static ssize_t sn_force_interrupt_write_proc(struct file *file, - const char __user *buffer, size_t count, loff_t *data) -{ - char val; - - if (copy_from_user(&val, buffer, 1)) - return -EFAULT; - - sn_force_interrupt_flag = (val == '0') ? 0 : 1; - return count; -} - -static int sn_force_interrupt_open(struct inode *inode, struct file *file) -{ - return single_open(file, sn_force_interrupt_show, NULL); -} - static int coherence_id_show(struct seq_file *s, void *p) { seq_printf(s, "%d\n", partition_coherence_id()); @@ -89,35 +57,45 @@ static int coherence_id_open(struct inode *inode, struct file *file) return single_open(file, coherence_id_show, NULL); } -static struct proc_dir_entry -*sn_procfs_create_entry(const char *name, struct proc_dir_entry *parent, - int (*openfunc)(struct inode *, struct file *), - int (*releasefunc)(struct inode *, struct file *), - ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *)) -{ - struct proc_dir_entry *e = create_proc_entry(name, 0444, parent); - - if (e) { - struct file_operations *f; - - f = kzalloc(sizeof(*f), GFP_KERNEL); - if (f) { - f->open = openfunc; - f->read = seq_read; - f->llseek = seq_lseek; - f->release = releasefunc; - f->write = write; - e->proc_fops = f; - } - } - - return e; -} - /* /proc/sgi_sn/sn_topology uses seq_file, see sn_hwperf.c */ extern int sn_topology_open(struct inode *, struct file *); extern int sn_topology_release(struct inode *, struct file *); +static const struct file_operations proc_partition_id_fops = { + .open = partition_id_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static const struct file_operations proc_system_sn_fops = { + .open = system_serial_number_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static const struct file_operations proc_license_id_fops = { + .open = licenseID_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static const struct file_operations proc_coherence_id_fops = { + .open = coherence_id_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static const struct file_operations proc_sn_topo_fops = { + .open = sn_topology_open, + .read = seq_read, + .llseek = seq_lseek, + .release = sn_topology_release, +}; + void register_sn_procfs(void) { static struct proc_dir_entry *sgi_proc_dir = NULL; @@ -126,24 +104,14 @@ void register_sn_procfs(void) if (!(sgi_proc_dir = proc_mkdir("sgi_sn", NULL))) return; - sn_procfs_create_entry("partition_id", sgi_proc_dir, - partition_id_open, single_release, NULL); - - sn_procfs_create_entry("system_serial_number", sgi_proc_dir, - system_serial_number_open, single_release, NULL); - - sn_procfs_create_entry("licenseID", sgi_proc_dir, - licenseID_open, single_release, NULL); - - sn_procfs_create_entry("sn_force_interrupt", sgi_proc_dir, - sn_force_interrupt_open, single_release, - sn_force_interrupt_write_proc); - - sn_procfs_create_entry("coherence_id", sgi_proc_dir, - coherence_id_open, single_release, NULL); - - sn_procfs_create_entry("sn_topology", sgi_proc_dir, - sn_topology_open, sn_topology_release, NULL); + proc_create("partition_id", 0444, sgi_proc_dir, + &proc_partition_id_fops); + proc_create("system_serial_number", 0444, sgi_proc_dir, + &proc_system_sn_fops); + proc_create("licenseID", 0444, sgi_proc_dir, &proc_license_id_fops); + proc_create("coherence_id", 0444, sgi_proc_dir, + &proc_coherence_id_fops); + proc_create("sn_topology", 0444, sgi_proc_dir, &proc_sn_topo_fops); } #endif /* CONFIG_PROC_FS */ diff --git a/arch/ia64/sn/kernel/sn2/timer.c b/arch/ia64/sn/kernel/sn2/timer.c index 56a88b6df4b..abab8f99e91 100644 --- a/arch/ia64/sn/kernel/sn2/timer.c +++ b/arch/ia64/sn/kernel/sn2/timer.c @@ -11,9 +11,9 @@ #include <linux/sched.h> #include <linux/time.h> #include <linux/interrupt.h> +#include <linux/clocksource.h> #include <asm/hw_irq.h> -#include <asm/system.h> #include <asm/timex.h> #include <asm/sn/leds.h> @@ -22,11 +22,17 @@ extern unsigned long sn_rtc_cycles_per_second; -static struct time_interpolator sn2_interpolator = { - .drift = -1, - .shift = 10, - .mask = (1LL << 55) - 1, - .source = TIME_SOURCE_MMIO64 +static cycle_t read_sn2(struct clocksource *cs) +{ + return (cycle_t)readq(RTC_COUNTER_ADDR); +} + +static struct clocksource clocksource_sn2 = { + .name = "sn2_rtc", + .rating = 450, + .read = read_sn2, + .mask = (1LL << 55) - 1, + .flags = CLOCK_SOURCE_IS_CONTINUOUS, }; /* @@ -47,9 +53,8 @@ ia64_sn_udelay (unsigned long usecs) void __init sn_timer_init(void) { - sn2_interpolator.frequency = sn_rtc_cycles_per_second; - sn2_interpolator.addr = RTC_COUNTER_ADDR; - register_time_interpolator(&sn2_interpolator); + clocksource_sn2.archdata.fsys_mmio = RTC_COUNTER_ADDR; + clocksource_register_hz(&clocksource_sn2, sn_rtc_cycles_per_second); ia64_udelay = &ia64_sn_udelay; } diff --git a/arch/ia64/sn/kernel/sn2/timer_interrupt.c b/arch/ia64/sn/kernel/sn2/timer_interrupt.c index fa7f6994591..103d6ea8e94 100644 --- a/arch/ia64/sn/kernel/sn2/timer_interrupt.c +++ b/arch/ia64/sn/kernel/sn2/timer_interrupt.c @@ -36,7 +36,7 @@ extern irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs); #define SN_LB_INT_WAR_INTERVAL 100 -void sn_timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) +void sn_timer_interrupt(int irq, void *dev_id) { /* LED blinking */ if (!pda->hb_count--) { diff --git a/arch/ia64/sn/kernel/tiocx.c b/arch/ia64/sn/kernel/tiocx.c index feaf1a6e810..e35f6485c1f 100644 --- a/arch/ia64/sn/kernel/tiocx.c +++ b/arch/ia64/sn/kernel/tiocx.c @@ -14,7 +14,6 @@ #include <linux/capability.h> #include <linux/device.h> #include <linux/delay.h> -#include <asm/system.h> #include <asm/uaccess.h> #include <asm/sn/sn_sal.h> #include <asm/sn/addrs.h> @@ -66,8 +65,7 @@ static int tiocx_match(struct device *dev, struct device_driver *drv) } -static int tiocx_uevent(struct device *dev, char **envp, int num_envp, - char *buffer, int buffer_size) +static int tiocx_uevent(struct device *dev, struct kobj_uevent_env *env) { return -ENODEV; } @@ -192,6 +190,7 @@ cx_device_register(nasid_t nasid, int part_num, int mfg_num, struct hubdev_info *hubdev, int bt) { struct cx_dev *cx_dev; + int r; cx_dev = kzalloc(sizeof(struct cx_dev), GFP_KERNEL); DBG("cx_dev= 0x%p\n", cx_dev); @@ -207,9 +206,12 @@ cx_device_register(nasid_t nasid, int part_num, int mfg_num, cx_dev->dev.parent = NULL; cx_dev->dev.bus = &tiocx_bus_type; cx_dev->dev.release = tiocx_bus_release; - snprintf(cx_dev->dev.bus_id, BUS_ID_SIZE, "%d", - cx_dev->cx_id.nasid); - device_register(&cx_dev->dev); + dev_set_name(&cx_dev->dev, "%d", cx_dev->cx_id.nasid); + r = device_register(&cx_dev->dev); + if (r) { + kfree(cx_dev); + return r; + } get_device(&cx_dev->dev); device_create_file(&cx_dev->dev, &dev_attr_cxdev_control); @@ -369,8 +371,8 @@ static void tio_corelet_reset(nasid_t nasid, int corelet) static int is_fpga_tio(int nasid, int *bt) { - u16 ioboard_type; - s64 rc; + u16 uninitialized_var(ioboard_type); /* GCC be quiet */ + long rc; rc = ia64_sn_sysctl_ioboard_get(nasid, &ioboard_type); if (rc) { @@ -488,11 +490,14 @@ static int __init tiocx_init(void) { cnodeid_t cnodeid; int found_tiocx_device = 0; + int err; if (!ia64_platform_is("sn2")) return 0; - bus_register(&tiocx_bus_type); + err = bus_register(&tiocx_bus_type); + if (err) + return err; for (cnodeid = 0; cnodeid < num_cnodes; cnodeid++) { nasid_t nasid; @@ -552,7 +557,7 @@ static void __exit tiocx_exit(void) bus_unregister(&tiocx_bus_type); } -subsys_initcall(tiocx_init); +fs_initcall(tiocx_init); module_exit(tiocx_exit); /************************************************************************ diff --git a/arch/ia64/sn/kernel/xp_main.c b/arch/ia64/sn/kernel/xp_main.c deleted file mode 100644 index b7ea46645e1..00000000000 --- a/arch/ia64/sn/kernel/xp_main.c +++ /dev/null @@ -1,290 +0,0 @@ -/* - * This file is subject to the terms and conditions of the GNU General Public - * License. See the file "COPYING" in the main directory of this archive - * for more details. - * - * Copyright (c) 2004-2005 Silicon Graphics, Inc. All Rights Reserved. - */ - - -/* - * Cross Partition (XP) base. - * - * XP provides a base from which its users can interact - * with XPC, yet not be dependent on XPC. - * - */ - - -#include <linux/kernel.h> -#include <linux/interrupt.h> -#include <linux/module.h> -#include <linux/mutex.h> -#include <asm/sn/intr.h> -#include <asm/sn/sn_sal.h> -#include <asm/sn/xp.h> - - -/* - * Target of nofault PIO read. - */ -u64 xp_nofault_PIOR_target; - - -/* - * xpc_registrations[] keeps track of xpc_connect()'s done by the kernel-level - * users of XPC. - */ -struct xpc_registration xpc_registrations[XPC_NCHANNELS]; - - -/* - * Initialize the XPC interface to indicate that XPC isn't loaded. - */ -static enum xpc_retval xpc_notloaded(void) { return xpcNotLoaded; } - -struct xpc_interface xpc_interface = { - (void (*)(int)) xpc_notloaded, - (void (*)(int)) xpc_notloaded, - (enum xpc_retval (*)(partid_t, int, u32, void **)) xpc_notloaded, - (enum xpc_retval (*)(partid_t, int, void *)) xpc_notloaded, - (enum xpc_retval (*)(partid_t, int, void *, xpc_notify_func, void *)) - xpc_notloaded, - (void (*)(partid_t, int, void *)) xpc_notloaded, - (enum xpc_retval (*)(partid_t, void *)) xpc_notloaded -}; - - -/* - * XPC calls this when it (the XPC module) has been loaded. - */ -void -xpc_set_interface(void (*connect)(int), - void (*disconnect)(int), - enum xpc_retval (*allocate)(partid_t, int, u32, void **), - enum xpc_retval (*send)(partid_t, int, void *), - enum xpc_retval (*send_notify)(partid_t, int, void *, - xpc_notify_func, void *), - void (*received)(partid_t, int, void *), - enum xpc_retval (*partid_to_nasids)(partid_t, void *)) -{ - xpc_interface.connect = connect; - xpc_interface.disconnect = disconnect; - xpc_interface.allocate = allocate; - xpc_interface.send = send; - xpc_interface.send_notify = send_notify; - xpc_interface.received = received; - xpc_interface.partid_to_nasids = partid_to_nasids; -} - - -/* - * XPC calls this when it (the XPC module) is being unloaded. - */ -void -xpc_clear_interface(void) -{ - xpc_interface.connect = (void (*)(int)) xpc_notloaded; - xpc_interface.disconnect = (void (*)(int)) xpc_notloaded; - xpc_interface.allocate = (enum xpc_retval (*)(partid_t, int, u32, - void **)) xpc_notloaded; - xpc_interface.send = (enum xpc_retval (*)(partid_t, int, void *)) - xpc_notloaded; - xpc_interface.send_notify = (enum xpc_retval (*)(partid_t, int, void *, - xpc_notify_func, void *)) xpc_notloaded; - xpc_interface.received = (void (*)(partid_t, int, void *)) - xpc_notloaded; - xpc_interface.partid_to_nasids = (enum xpc_retval (*)(partid_t, void *)) - xpc_notloaded; -} - - -/* - * Register for automatic establishment of a channel connection whenever - * a partition comes up. - * - * Arguments: - * - * ch_number - channel # to register for connection. - * func - function to call for asynchronous notification of channel - * state changes (i.e., connection, disconnection, error) and - * the arrival of incoming messages. - * key - pointer to optional user-defined value that gets passed back - * to the user on any callouts made to func. - * payload_size - size in bytes of the XPC message's payload area which - * contains a user-defined message. The user should make - * this large enough to hold their largest message. - * nentries - max #of XPC message entries a message queue can contain. - * The actual number, which is determined when a connection - * is established and may be less then requested, will be - * passed to the user via the xpcConnected callout. - * assigned_limit - max number of kthreads allowed to be processing - * messages (per connection) at any given instant. - * idle_limit - max number of kthreads allowed to be idle at any given - * instant. - */ -enum xpc_retval -xpc_connect(int ch_number, xpc_channel_func func, void *key, u16 payload_size, - u16 nentries, u32 assigned_limit, u32 idle_limit) -{ - struct xpc_registration *registration; - - - DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS); - DBUG_ON(payload_size == 0 || nentries == 0); - DBUG_ON(func == NULL); - DBUG_ON(assigned_limit == 0 || idle_limit > assigned_limit); - - registration = &xpc_registrations[ch_number]; - - if (mutex_lock_interruptible(®istration->mutex) != 0) { - return xpcInterrupted; - } - - /* if XPC_CHANNEL_REGISTERED(ch_number) */ - if (registration->func != NULL) { - mutex_unlock(®istration->mutex); - return xpcAlreadyRegistered; - } - - /* register the channel for connection */ - registration->msg_size = XPC_MSG_SIZE(payload_size); - registration->nentries = nentries; - registration->assigned_limit = assigned_limit; - registration->idle_limit = idle_limit; - registration->key = key; - registration->func = func; - - mutex_unlock(®istration->mutex); - - xpc_interface.connect(ch_number); - - return xpcSuccess; -} - - -/* - * Remove the registration for automatic connection of the specified channel - * when a partition comes up. - * - * Before returning this xpc_disconnect() will wait for all connections on the - * specified channel have been closed/torndown. So the caller can be assured - * that they will not be receiving any more callouts from XPC to their - * function registered via xpc_connect(). - * - * Arguments: - * - * ch_number - channel # to unregister. - */ -void -xpc_disconnect(int ch_number) -{ - struct xpc_registration *registration; - - - DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS); - - registration = &xpc_registrations[ch_number]; - - /* - * We've decided not to make this a down_interruptible(), since we - * figured XPC's users will just turn around and call xpc_disconnect() - * again anyways, so we might as well wait, if need be. - */ - mutex_lock(®istration->mutex); - - /* if !XPC_CHANNEL_REGISTERED(ch_number) */ - if (registration->func == NULL) { - mutex_unlock(®istration->mutex); - return; - } - - /* remove the connection registration for the specified channel */ - registration->func = NULL; - registration->key = NULL; - registration->nentries = 0; - registration->msg_size = 0; - registration->assigned_limit = 0; - registration->idle_limit = 0; - - xpc_interface.disconnect(ch_number); - - mutex_unlock(®istration->mutex); - - return; -} - - -int __init -xp_init(void) -{ - int ret, ch_number; - u64 func_addr = *(u64 *) xp_nofault_PIOR; - u64 err_func_addr = *(u64 *) xp_error_PIOR; - - - if (!ia64_platform_is("sn2")) { - return -ENODEV; - } - - /* - * Register a nofault code region which performs a cross-partition - * PIO read. If the PIO read times out, the MCA handler will consume - * the error and return to a kernel-provided instruction to indicate - * an error. This PIO read exists because it is guaranteed to timeout - * if the destination is down (AMO operations do not timeout on at - * least some CPUs on Shubs <= v1.2, which unfortunately we have to - * work around). - */ - if ((ret = sn_register_nofault_code(func_addr, err_func_addr, - err_func_addr, 1, 1)) != 0) { - printk(KERN_ERR "XP: can't register nofault code, error=%d\n", - ret); - } - /* - * Setup the nofault PIO read target. (There is no special reason why - * SH_IPI_ACCESS was selected.) - */ - if (is_shub2()) { - xp_nofault_PIOR_target = SH2_IPI_ACCESS0; - } else { - xp_nofault_PIOR_target = SH1_IPI_ACCESS; - } - - /* initialize the connection registration mutex */ - for (ch_number = 0; ch_number < XPC_NCHANNELS; ch_number++) { - mutex_init(&xpc_registrations[ch_number].mutex); - } - - return 0; -} -module_init(xp_init); - - -void __exit -xp_exit(void) -{ - u64 func_addr = *(u64 *) xp_nofault_PIOR; - u64 err_func_addr = *(u64 *) xp_error_PIOR; - - - /* unregister the PIO read nofault code region */ - (void) sn_register_nofault_code(func_addr, err_func_addr, - err_func_addr, 1, 0); -} -module_exit(xp_exit); - - -MODULE_AUTHOR("Silicon Graphics, Inc."); -MODULE_DESCRIPTION("Cross Partition (XP) base"); -MODULE_LICENSE("GPL"); - -EXPORT_SYMBOL(xp_nofault_PIOR); -EXPORT_SYMBOL(xp_nofault_PIOR_target); -EXPORT_SYMBOL(xpc_registrations); -EXPORT_SYMBOL(xpc_interface); -EXPORT_SYMBOL(xpc_clear_interface); -EXPORT_SYMBOL(xpc_set_interface); -EXPORT_SYMBOL(xpc_connect); -EXPORT_SYMBOL(xpc_disconnect); - diff --git a/arch/ia64/sn/kernel/xp_nofault.S b/arch/ia64/sn/kernel/xp_nofault.S deleted file mode 100644 index b772543053c..00000000000 --- a/arch/ia64/sn/kernel/xp_nofault.S +++ /dev/null @@ -1,31 +0,0 @@ -/* - * This file is subject to the terms and conditions of the GNU General Public - * License. See the file "COPYING" in the main directory of this archive - * for more details. - * - * Copyright (c) 2004-2005 Silicon Graphics, Inc. All Rights Reserved. - */ - - -/* - * The xp_nofault_PIOR function takes a pointer to a remote PIO register - * and attempts to load and consume a value from it. This function - * will be registered as a nofault code block. In the event that the - * PIO read fails, the MCA handler will force the error to look - * corrected and vector to the xp_error_PIOR which will return an error. - * - * extern int xp_nofault_PIOR(void *remote_register); - */ - - .global xp_nofault_PIOR -xp_nofault_PIOR: - mov r8=r0 // Stage a success return value - ld8.acq r9=[r32];; // PIO Read the specified register - adds r9=1,r9 // Add to force a consume - br.ret.sptk.many b0;; // Return success - - .global xp_error_PIOR -xp_error_PIOR: - mov r8=1 // Return value of 1 - br.ret.sptk.many b0;; // Return failure - diff --git a/arch/ia64/sn/kernel/xpc_channel.c b/arch/ia64/sn/kernel/xpc_channel.c deleted file mode 100644 index 1f3540826e6..00000000000 --- a/arch/ia64/sn/kernel/xpc_channel.c +++ /dev/null @@ -1,2374 +0,0 @@ -/* - * This file is subject to the terms and conditions of the GNU General Public - * License. See the file "COPYING" in the main directory of this archive - * for more details. - * - * Copyright (c) 2004-2006 Silicon Graphics, Inc. All Rights Reserved. - */ - - -/* - * Cross Partition Communication (XPC) channel support. - * - * This is the part of XPC that manages the channels and - * sends/receives messages across them to/from other partitions. - * - */ - - -#include <linux/kernel.h> -#include <linux/init.h> -#include <linux/sched.h> -#include <linux/cache.h> -#include <linux/interrupt.h> -#include <linux/mutex.h> -#include <linux/completion.h> -#include <asm/sn/bte.h> -#include <asm/sn/sn_sal.h> -#include <asm/sn/xpc.h> - - -/* - * Guarantee that the kzalloc'd memory is cacheline aligned. - */ -static void * -xpc_kzalloc_cacheline_aligned(size_t size, gfp_t flags, void **base) -{ - /* see if kzalloc will give us cachline aligned memory by default */ - *base = kzalloc(size, flags); - if (*base == NULL) { - return NULL; - } - if ((u64) *base == L1_CACHE_ALIGN((u64) *base)) { - return *base; - } - kfree(*base); - - /* nope, we'll have to do it ourselves */ - *base = kzalloc(size + L1_CACHE_BYTES, flags); - if (*base == NULL) { - return NULL; - } - return (void *) L1_CACHE_ALIGN((u64) *base); -} - - -/* - * Set up the initial values for the XPartition Communication channels. - */ -static void -xpc_initialize_channels(struct xpc_partition *part, partid_t partid) -{ - int ch_number; - struct xpc_channel *ch; - - - for (ch_number = 0; ch_number < part->nchannels; ch_number++) { - ch = &part->channels[ch_number]; - - ch->partid = partid; - ch->number = ch_number; - ch->flags = XPC_C_DISCONNECTED; - - ch->local_GP = &part->local_GPs[ch_number]; - ch->local_openclose_args = - &part->local_openclose_args[ch_number]; - - atomic_set(&ch->kthreads_assigned, 0); - atomic_set(&ch->kthreads_idle, 0); - atomic_set(&ch->kthreads_active, 0); - - atomic_set(&ch->references, 0); - atomic_set(&ch->n_to_notify, 0); - - spin_lock_init(&ch->lock); - mutex_init(&ch->msg_to_pull_mutex); - init_completion(&ch->wdisconnect_wait); - - atomic_set(&ch->n_on_msg_allocate_wq, 0); - init_waitqueue_head(&ch->msg_allocate_wq); - init_waitqueue_head(&ch->idle_wq); - } -} - - -/* - * Setup the infrastructure necessary to support XPartition Communication - * between the specified remote partition and the local one. - */ -enum xpc_retval -xpc_setup_infrastructure(struct xpc_partition *part) -{ - int ret, cpuid; - struct timer_list *timer; - partid_t partid = XPC_PARTID(part); - - - /* - * Zero out MOST of the entry for this partition. Only the fields - * starting with `nchannels' will be zeroed. The preceding fields must - * remain `viable' across partition ups and downs, since they may be - * referenced during this memset() operation. - */ - memset(&part->nchannels, 0, sizeof(struct xpc_partition) - - offsetof(struct xpc_partition, nchannels)); - - /* - * Allocate all of the channel structures as a contiguous chunk of - * memory. - */ - part->channels = kzalloc(sizeof(struct xpc_channel) * XPC_NCHANNELS, - GFP_KERNEL); - if (part->channels == NULL) { - dev_err(xpc_chan, "can't get memory for channels\n"); - return xpcNoMemory; - } - - part->nchannels = XPC_NCHANNELS; - - - /* allocate all the required GET/PUT values */ - - part->local_GPs = xpc_kzalloc_cacheline_aligned(XPC_GP_SIZE, - GFP_KERNEL, &part->local_GPs_base); - if (part->local_GPs == NULL) { - kfree(part->channels); - part->channels = NULL; - dev_err(xpc_chan, "can't get memory for local get/put " - "values\n"); - return xpcNoMemory; - } - - part->remote_GPs = xpc_kzalloc_cacheline_aligned(XPC_GP_SIZE, - GFP_KERNEL, &part->remote_GPs_base); - if (part->remote_GPs == NULL) { - dev_err(xpc_chan, "can't get memory for remote get/put " - "values\n"); - kfree(part->local_GPs_base); - part->local_GPs = NULL; - kfree(part->channels); - part->channels = NULL; - return xpcNoMemory; - } - - - /* allocate all the required open and close args */ - - part->local_openclose_args = xpc_kzalloc_cacheline_aligned( - XPC_OPENCLOSE_ARGS_SIZE, GFP_KERNEL, - &part->local_openclose_args_base); - if (part->local_openclose_args == NULL) { - dev_err(xpc_chan, "can't get memory for local connect args\n"); - kfree(part->remote_GPs_base); - part->remote_GPs = NULL; - kfree(part->local_GPs_base); - part->local_GPs = NULL; - kfree(part->channels); - part->channels = NULL; - return xpcNoMemory; - } - - part->remote_openclose_args = xpc_kzalloc_cacheline_aligned( - XPC_OPENCLOSE_ARGS_SIZE, GFP_KERNEL, - &part->remote_openclose_args_base); - if (part->remote_openclose_args == NULL) { - dev_err(xpc_chan, "can't get memory for remote connect args\n"); - kfree(part->local_openclose_args_base); - part->local_openclose_args = NULL; - kfree(part->remote_GPs_base); - part->remote_GPs = NULL; - kfree(part->local_GPs_base); - part->local_GPs = NULL; - kfree(part->channels); - part->channels = NULL; - return xpcNoMemory; - } - - - xpc_initialize_channels(part, partid); - - atomic_set(&part->nchannels_active, 0); - atomic_set(&part->nchannels_engaged, 0); - - - /* local_IPI_amo were set to 0 by an earlier memset() */ - - /* Initialize this partitions AMO_t structure */ - part->local_IPI_amo_va = xpc_IPI_init(partid); - - spin_lock_init(&part->IPI_lock); - - atomic_set(&part->channel_mgr_requests, 1); - init_waitqueue_head(&part->channel_mgr_wq); - - sprintf(part->IPI_owner, "xpc%02d", partid); - ret = request_irq(SGI_XPC_NOTIFY, xpc_notify_IRQ_handler, IRQF_SHARED, - part->IPI_owner, (void *) (u64) partid); - if (ret != 0) { - dev_err(xpc_chan, "can't register NOTIFY IRQ handler, " - "errno=%d\n", -ret); - kfree(part->remote_openclose_args_base); - part->remote_openclose_args = NULL; - kfree(part->local_openclose_args_base); - part->local_openclose_args = NULL; - kfree(part->remote_GPs_base); - part->remote_GPs = NULL; - kfree(part->local_GPs_base); - part->local_GPs = NULL; - kfree(part->channels); - part->channels = NULL; - return xpcLackOfResources; - } - - /* Setup a timer to check for dropped IPIs */ - timer = &part->dropped_IPI_timer; - init_timer(timer); - timer->function = (void (*)(unsigned long)) xpc_dropped_IPI_check; - timer->data = (unsigned long) part; - timer->expires = jiffies + XPC_P_DROPPED_IPI_WAIT; - add_timer(timer); - - /* - * With the setting of the partition setup_state to XPC_P_SETUP, we're - * declaring that this partition is ready to go. - */ - part->setup_state = XPC_P_SETUP; - - - /* - * Setup the per partition specific variables required by the - * remote partition to establish channel connections with us. - * - * The setting of the magic # indicates that these per partition - * specific variables are ready to be used. - */ - xpc_vars_part[partid].GPs_pa = __pa(part->local_GPs); - xpc_vars_part[partid].openclose_args_pa = - __pa(part->local_openclose_args); - xpc_vars_part[partid].IPI_amo_pa = __pa(part->local_IPI_amo_va); - cpuid = raw_smp_processor_id(); /* any CPU in this partition will do */ - xpc_vars_part[partid].IPI_nasid = cpuid_to_nasid(cpuid); - xpc_vars_part[partid].IPI_phys_cpuid = cpu_physical_id(cpuid); - xpc_vars_part[partid].nchannels = part->nchannels; - xpc_vars_part[partid].magic = XPC_VP_MAGIC1; - - return xpcSuccess; -} - - -/* - * Create a wrapper that hides the underlying mechanism for pulling a cacheline - * (or multiple cachelines) from a remote partition. - * - * src must be a cacheline aligned physical address on the remote partition. - * dst must be a cacheline aligned virtual address on this partition. - * cnt must be an cacheline sized - */ -static enum xpc_retval -xpc_pull_remote_cachelines(struct xpc_partition *part, void *dst, - const void *src, size_t cnt) -{ - bte_result_t bte_ret; - - - DBUG_ON((u64) src != L1_CACHE_ALIGN((u64) src)); - DBUG_ON((u64) dst != L1_CACHE_ALIGN((u64) dst)); - DBUG_ON(cnt != L1_CACHE_ALIGN(cnt)); - - if (part->act_state == XPC_P_DEACTIVATING) { - return part->reason; - } - - bte_ret = xp_bte_copy((u64) src, (u64) dst, (u64) cnt, - (BTE_NORMAL | BTE_WACQUIRE), NULL); - if (bte_ret == BTE_SUCCESS) { - return xpcSuccess; - } - - dev_dbg(xpc_chan, "xp_bte_copy() from partition %d failed, ret=%d\n", - XPC_PARTID(part), bte_ret); - - return xpc_map_bte_errors(bte_ret); -} - - -/* - * Pull the remote per partititon specific variables from the specified - * partition. - */ -enum xpc_retval -xpc_pull_remote_vars_part(struct xpc_partition *part) -{ - u8 buffer[L1_CACHE_BYTES * 2]; - struct xpc_vars_part *pulled_entry_cacheline = - (struct xpc_vars_part *) L1_CACHE_ALIGN((u64) buffer); - struct xpc_vars_part *pulled_entry; - u64 remote_entry_cacheline_pa, remote_entry_pa; - partid_t partid = XPC_PARTID(part); - enum xpc_retval ret; - - - /* pull the cacheline that contains the variables we're interested in */ - - DBUG_ON(part->remote_vars_part_pa != - L1_CACHE_ALIGN(part->remote_vars_part_pa)); - DBUG_ON(sizeof(struct xpc_vars_part) != L1_CACHE_BYTES / 2); - - remote_entry_pa = part->remote_vars_part_pa + - sn_partition_id * sizeof(struct xpc_vars_part); - - remote_entry_cacheline_pa = (remote_entry_pa & ~(L1_CACHE_BYTES - 1)); - - pulled_entry = (struct xpc_vars_part *) ((u64) pulled_entry_cacheline + - (remote_entry_pa & (L1_CACHE_BYTES - 1))); - - ret = xpc_pull_remote_cachelines(part, pulled_entry_cacheline, - (void *) remote_entry_cacheline_pa, - L1_CACHE_BYTES); - if (ret != xpcSuccess) { - dev_dbg(xpc_chan, "failed to pull XPC vars_part from " - "partition %d, ret=%d\n", partid, ret); - return ret; - } - - - /* see if they've been set up yet */ - - if (pulled_entry->magic != XPC_VP_MAGIC1 && - pulled_entry->magic != XPC_VP_MAGIC2) { - - if (pulled_entry->magic != 0) { - dev_dbg(xpc_chan, "partition %d's XPC vars_part for " - "partition %d has bad magic value (=0x%lx)\n", - partid, sn_partition_id, pulled_entry->magic); - return xpcBadMagic; - } - - /* they've not been initialized yet */ - return xpcRetry; - } - - if (xpc_vars_part[partid].magic == XPC_VP_MAGIC1) { - - /* validate the variables */ - - if (pulled_entry->GPs_pa == 0 || - pulled_entry->openclose_args_pa == 0 || - pulled_entry->IPI_amo_pa == 0) { - - dev_err(xpc_chan, "partition %d's XPC vars_part for " - "partition %d are not valid\n", partid, - sn_partition_id); - return xpcInvalidAddress; - } - - /* the variables we imported look to be valid */ - - part->remote_GPs_pa = pulled_entry->GPs_pa; - part->remote_openclose_args_pa = - pulled_entry->openclose_args_pa; - part->remote_IPI_amo_va = - (AMO_t *) __va(pulled_entry->IPI_amo_pa); - part->remote_IPI_nasid = pulled_entry->IPI_nasid; - part->remote_IPI_phys_cpuid = pulled_entry->IPI_phys_cpuid; - - if (part->nchannels > pulled_entry->nchannels) { - part->nchannels = pulled_entry->nchannels; - } - - /* let the other side know that we've pulled their variables */ - - xpc_vars_part[partid].magic = XPC_VP_MAGIC2; - } - - if (pulled_entry->magic == XPC_VP_MAGIC1) { - return xpcRetry; - } - - return xpcSuccess; -} - - -/* - * Get the IPI flags and pull the openclose args and/or remote GPs as needed. - */ -static u64 -xpc_get_IPI_flags(struct xpc_partition *part) -{ - unsigned long irq_flags; - u64 IPI_amo; - enum xpc_retval ret; - - - /* - * See if there are any IPI flags to be handled. - */ - - spin_lock_irqsave(&part->IPI_lock, irq_flags); - if ((IPI_amo = part->local_IPI_amo) != 0) { - part->local_IPI_amo = 0; - } - spin_unlock_irqrestore(&part->IPI_lock, irq_flags); - - - if (XPC_ANY_OPENCLOSE_IPI_FLAGS_SET(IPI_amo)) { - ret = xpc_pull_remote_cachelines(part, - part->remote_openclose_args, - (void *) part->remote_openclose_args_pa, - XPC_OPENCLOSE_ARGS_SIZE); - if (ret != xpcSuccess) { - XPC_DEACTIVATE_PARTITION(part, ret); - - dev_dbg(xpc_chan, "failed to pull openclose args from " - "partition %d, ret=%d\n", XPC_PARTID(part), - ret); - - /* don't bother processing IPIs anymore */ - IPI_amo = 0; - } - } - - if (XPC_ANY_MSG_IPI_FLAGS_SET(IPI_amo)) { - ret = xpc_pull_remote_cachelines(part, part->remote_GPs, - (void *) part->remote_GPs_pa, - XPC_GP_SIZE); - if (ret != xpcSuccess) { - XPC_DEACTIVATE_PARTITION(part, ret); - - dev_dbg(xpc_chan, "failed to pull GPs from partition " - "%d, ret=%d\n", XPC_PARTID(part), ret); - - /* don't bother processing IPIs anymore */ - IPI_amo = 0; - } - } - - return IPI_amo; -} - - -/* - * Allocate the local message queue and the notify queue. - */ -static enum xpc_retval -xpc_allocate_local_msgqueue(struct xpc_channel *ch) -{ - unsigned long irq_flags; - int nentries; - size_t nbytes; - - - // >>> may want to check for ch->flags & XPC_C_DISCONNECTING between - // >>> iterations of the for-loop, bail if set? - - // >>> should we impose a minumum #of entries? like 4 or 8? - for (nentries = ch->local_nentries; nentries > 0; nentries--) { - - nbytes = nentries * ch->msg_size; - ch->local_msgqueue = xpc_kzalloc_cacheline_aligned(nbytes, - GFP_KERNEL, - &ch->local_msgqueue_base); - if (ch->local_msgqueue == NULL) { - continue; - } - - nbytes = nentries * sizeof(struct xpc_notify); - ch->notify_queue = kzalloc(nbytes, GFP_KERNEL); - if (ch->notify_queue == NULL) { - kfree(ch->local_msgqueue_base); - ch->local_msgqueue = NULL; - continue; - } - - spin_lock_irqsave(&ch->lock, irq_flags); - if (nentries < ch->local_nentries) { - dev_dbg(xpc_chan, "nentries=%d local_nentries=%d, " - "partid=%d, channel=%d\n", nentries, - ch->local_nentries, ch->partid, ch->number); - - ch->local_nentries = nentries; - } - spin_unlock_irqrestore(&ch->lock, irq_flags); - return xpcSuccess; - } - - dev_dbg(xpc_chan, "can't get memory for local message queue and notify " - "queue, partid=%d, channel=%d\n", ch->partid, ch->number); - return xpcNoMemory; -} - - -/* - * Allocate the cached remote message queue. - */ -static enum xpc_retval -xpc_allocate_remote_msgqueue(struct xpc_channel *ch) -{ - unsigned long irq_flags; - int nentries; - size_t nbytes; - - - DBUG_ON(ch->remote_nentries <= 0); - - // >>> may want to check for ch->flags & XPC_C_DISCONNECTING between - // >>> iterations of the for-loop, bail if set? - - // >>> should we impose a minumum #of entries? like 4 or 8? - for (nentries = ch->remote_nentries; nentries > 0; nentries--) { - - nbytes = nentries * ch->msg_size; - ch->remote_msgqueue = xpc_kzalloc_cacheline_aligned(nbytes, - GFP_KERNEL, - &ch->remote_msgqueue_base); - if (ch->remote_msgqueue == NULL) { - continue; - } - - spin_lock_irqsave(&ch->lock, irq_flags); - if (nentries < ch->remote_nentries) { - dev_dbg(xpc_chan, "nentries=%d remote_nentries=%d, " - "partid=%d, channel=%d\n", nentries, - ch->remote_nentries, ch->partid, ch->number); - - ch->remote_nentries = nentries; - } - spin_unlock_irqrestore(&ch->lock, irq_flags); - return xpcSuccess; - } - - dev_dbg(xpc_chan, "can't get memory for cached remote message queue, " - "partid=%d, channel=%d\n", ch->partid, ch->number); - return xpcNoMemory; -} - - -/* - * Allocate message queues and other stuff associated with a channel. - * - * Note: Assumes all of the channel sizes are filled in. - */ -static enum xpc_retval -xpc_allocate_msgqueues(struct xpc_channel *ch) -{ - unsigned long irq_flags; - enum xpc_retval ret; - - - DBUG_ON(ch->flags & XPC_C_SETUP); - - if ((ret = xpc_allocate_local_msgqueue(ch)) != xpcSuccess) { - return ret; - } - - if ((ret = xpc_allocate_remote_msgqueue(ch)) != xpcSuccess) { - kfree(ch->local_msgqueue_base); - ch->local_msgqueue = NULL; - kfree(ch->notify_queue); - ch->notify_queue = NULL; - return ret; - } - - spin_lock_irqsave(&ch->lock, irq_flags); - ch->flags |= XPC_C_SETUP; - spin_unlock_irqrestore(&ch->lock, irq_flags); - - return xpcSuccess; -} - - -/* - * Process a connect message from a remote partition. - * - * Note: xpc_process_connect() is expecting to be called with the - * spin_lock_irqsave held and will leave it locked upon return. - */ -static void -xpc_process_connect(struct xpc_channel *ch, unsigned long *irq_flags) -{ - enum xpc_retval ret; - - - DBUG_ON(!spin_is_locked(&ch->lock)); - - if (!(ch->flags & XPC_C_OPENREQUEST) || - !(ch->flags & XPC_C_ROPENREQUEST)) { - /* nothing more to do for now */ - return; - } - DBUG_ON(!(ch->flags & XPC_C_CONNECTING)); - - if (!(ch->flags & XPC_C_SETUP)) { - spin_unlock_irqrestore(&ch->lock, *irq_flags); - ret = xpc_allocate_msgqueues(ch); - spin_lock_irqsave(&ch->lock, *irq_flags); - - if (ret != xpcSuccess) { - XPC_DISCONNECT_CHANNEL(ch, ret, irq_flags); - } - if (ch->flags & (XPC_C_CONNECTED | XPC_C_DISCONNECTING)) { - return; - } - - DBUG_ON(!(ch->flags & XPC_C_SETUP)); - DBUG_ON(ch->local_msgqueue == NULL); - DBUG_ON(ch->remote_msgqueue == NULL); - } - - if (!(ch->flags & XPC_C_OPENREPLY)) { - ch->flags |= XPC_C_OPENREPLY; - xpc_IPI_send_openreply(ch, irq_flags); - } - - if (!(ch->flags & XPC_C_ROPENREPLY)) { - return; - } - - DBUG_ON(ch->remote_msgqueue_pa == 0); - - ch->flags = (XPC_C_CONNECTED | XPC_C_SETUP); /* clear all else */ - - dev_info(xpc_chan, "channel %d to partition %d connected\n", - ch->number, ch->partid); - - spin_unlock_irqrestore(&ch->lock, *irq_flags); - xpc_create_kthreads(ch, 1); - spin_lock_irqsave(&ch->lock, *irq_flags); -} - - -/* - * Notify those who wanted to be notified upon delivery of their message. - */ -static void -xpc_notify_senders(struct xpc_channel *ch, enum xpc_retval reason, s64 put) -{ - struct xpc_notify *notify; - u8 notify_type; - s64 get = ch->w_remote_GP.get - 1; - - - while (++get < put && atomic_read(&ch->n_to_notify) > 0) { - - notify = &ch->notify_queue[get % ch->local_nentries]; - - /* - * See if the notify entry indicates it was associated with - * a message who's sender wants to be notified. It is possible - * that it is, but someone else is doing or has done the - * notification. - */ - notify_type = notify->type; - if (notify_type == 0 || - cmpxchg(¬ify->type, notify_type, 0) != - notify_type) { - continue; - } - - DBUG_ON(notify_type != XPC_N_CALL); - - atomic_dec(&ch->n_to_notify); - - if (notify->func != NULL) { - dev_dbg(xpc_chan, "notify->func() called, notify=0x%p, " - "msg_number=%ld, partid=%d, channel=%d\n", - (void *) notify, get, ch->partid, ch->number); - - notify->func(reason, ch->partid, ch->number, - notify->key); - - dev_dbg(xpc_chan, "notify->func() returned, " - "notify=0x%p, msg_number=%ld, partid=%d, " - "channel=%d\n", (void *) notify, get, - ch->partid, ch->number); - } - } -} - - -/* - * Free up message queues and other stuff that were allocated for the specified - * channel. - * - * Note: ch->reason and ch->reason_line are left set for debugging purposes, - * they're cleared when XPC_C_DISCONNECTED is cleared. - */ -static void -xpc_free_msgqueues(struct xpc_channel *ch) -{ - DBUG_ON(!spin_is_locked(&ch->lock)); - DBUG_ON(atomic_read(&ch->n_to_notify) != 0); - - ch->remote_msgqueue_pa = 0; - ch->func = NULL; - ch->key = NULL; - ch->msg_size = 0; - ch->local_nentries = 0; - ch->remote_nentries = 0; - ch->kthreads_assigned_limit = 0; - ch->kthreads_idle_limit = 0; - - ch->local_GP->get = 0; - ch->local_GP->put = 0; - ch->remote_GP.get = 0; - ch->remote_GP.put = 0; - ch->w_local_GP.get = 0; - ch->w_local_GP.put = 0; - ch->w_remote_GP.get = 0; - ch->w_remote_GP.put = 0; - ch->next_msg_to_pull = 0; - - if (ch->flags & XPC_C_SETUP) { - ch->flags &= ~XPC_C_SETUP; - - dev_dbg(xpc_chan, "ch->flags=0x%x, partid=%d, channel=%d\n", - ch->flags, ch->partid, ch->number); - - kfree(ch->local_msgqueue_base); - ch->local_msgqueue = NULL; - kfree(ch->remote_msgqueue_base); - ch->remote_msgqueue = NULL; - kfree(ch->notify_queue); - ch->notify_queue = NULL; - } -} - - -/* - * spin_lock_irqsave() is expected to be held on entry. - */ -static void -xpc_process_disconnect(struct xpc_channel *ch, unsigned long *irq_flags) -{ - struct xpc_partition *part = &xpc_partitions[ch->partid]; - u32 channel_was_connected = (ch->flags & XPC_C_WASCONNECTED); - - - DBUG_ON(!spin_is_locked(&ch->lock)); - - if (!(ch->flags & XPC_C_DISCONNECTING)) { - return; - } - - DBUG_ON(!(ch->flags & XPC_C_CLOSEREQUEST)); - - /* make sure all activity has settled down first */ - - if (atomic_read(&ch->references) > 0 || - ((ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) && - !(ch->flags & XPC_C_DISCONNECTINGCALLOUT_MADE))) { - return; - } - DBUG_ON(atomic_read(&ch->kthreads_assigned) != 0); - - if (part->act_state == XPC_P_DEACTIVATING) { - /* can't proceed until the other side disengages from us */ - if (xpc_partition_engaged(1UL << ch->partid)) { - return; - } - - } else { - - /* as long as the other side is up do the full protocol */ - - if (!(ch->flags & XPC_C_RCLOSEREQUEST)) { - return; - } - - if (!(ch->flags & XPC_C_CLOSEREPLY)) { - ch->flags |= XPC_C_CLOSEREPLY; - xpc_IPI_send_closereply(ch, irq_flags); - } - - if (!(ch->flags & XPC_C_RCLOSEREPLY)) { - return; - } - } - - /* wake those waiting for notify completion */ - if (atomic_read(&ch->n_to_notify) > 0) { - /* >>> we do callout while holding ch->lock */ - xpc_notify_senders(ch, ch->reason, ch->w_local_GP.put); - } - - /* both sides are disconnected now */ - - if (ch->flags & XPC_C_DISCONNECTINGCALLOUT_MADE) { - spin_unlock_irqrestore(&ch->lock, *irq_flags); - xpc_disconnect_callout(ch, xpcDisconnected); - spin_lock_irqsave(&ch->lock, *irq_flags); - } - - /* it's now safe to free the channel's message queues */ - xpc_free_msgqueues(ch); - - /* mark disconnected, clear all other flags except XPC_C_WDISCONNECT */ - ch->flags = (XPC_C_DISCONNECTED | (ch->flags & XPC_C_WDISCONNECT)); - - atomic_dec(&part->nchannels_active); - - if (channel_was_connected) { - dev_info(xpc_chan, "channel %d to partition %d disconnected, " - "reason=%d\n", ch->number, ch->partid, ch->reason); - } - - if (ch->flags & XPC_C_WDISCONNECT) { - /* we won't lose the CPU since we're holding ch->lock */ - complete(&ch->wdisconnect_wait); - } else if (ch->delayed_IPI_flags) { - if (part->act_state != XPC_P_DEACTIVATING) { - /* time to take action on any delayed IPI flags */ - spin_lock(&part->IPI_lock); - XPC_SET_IPI_FLAGS(part->local_IPI_amo, ch->number, - ch->delayed_IPI_flags); - spin_unlock(&part->IPI_lock); - } - ch->delayed_IPI_flags = 0; - } -} - - -/* - * Process a change in the channel's remote connection state. - */ -static void -xpc_process_openclose_IPI(struct xpc_partition *part, int ch_number, - u8 IPI_flags) -{ - unsigned long irq_flags; - struct xpc_openclose_args *args = - &part->remote_openclose_args[ch_number]; - struct xpc_channel *ch = &part->channels[ch_number]; - enum xpc_retval reason; - - - - spin_lock_irqsave(&ch->lock, irq_flags); - -again: - - if ((ch->flags & XPC_C_DISCONNECTED) && - (ch->flags & XPC_C_WDISCONNECT)) { - /* - * Delay processing IPI flags until thread waiting disconnect - * has had a chance to see that the channel is disconnected. - */ - ch->delayed_IPI_flags |= IPI_flags; - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - - if (IPI_flags & XPC_IPI_CLOSEREQUEST) { - - dev_dbg(xpc_chan, "XPC_IPI_CLOSEREQUEST (reason=%d) received " - "from partid=%d, channel=%d\n", args->reason, - ch->partid, ch->number); - - /* - * If RCLOSEREQUEST is set, we're probably waiting for - * RCLOSEREPLY. We should find it and a ROPENREQUEST packed - * with this RCLOSEREQUEST in the IPI_flags. - */ - - if (ch->flags & XPC_C_RCLOSEREQUEST) { - DBUG_ON(!(ch->flags & XPC_C_DISCONNECTING)); - DBUG_ON(!(ch->flags & XPC_C_CLOSEREQUEST)); - DBUG_ON(!(ch->flags & XPC_C_CLOSEREPLY)); - DBUG_ON(ch->flags & XPC_C_RCLOSEREPLY); - - DBUG_ON(!(IPI_flags & XPC_IPI_CLOSEREPLY)); - IPI_flags &= ~XPC_IPI_CLOSEREPLY; - ch->flags |= XPC_C_RCLOSEREPLY; - - /* both sides have finished disconnecting */ - xpc_process_disconnect(ch, &irq_flags); - DBUG_ON(!(ch->flags & XPC_C_DISCONNECTED)); - goto again; - } - - if (ch->flags & XPC_C_DISCONNECTED) { - if (!(IPI_flags & XPC_IPI_OPENREQUEST)) { - if ((XPC_GET_IPI_FLAGS(part->local_IPI_amo, - ch_number) & XPC_IPI_OPENREQUEST)) { - - DBUG_ON(ch->delayed_IPI_flags != 0); - spin_lock(&part->IPI_lock); - XPC_SET_IPI_FLAGS(part->local_IPI_amo, - ch_number, - XPC_IPI_CLOSEREQUEST); - spin_unlock(&part->IPI_lock); - } - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - XPC_SET_REASON(ch, 0, 0); - ch->flags &= ~XPC_C_DISCONNECTED; - - atomic_inc(&part->nchannels_active); - ch->flags |= (XPC_C_CONNECTING | XPC_C_ROPENREQUEST); - } - - IPI_flags &= ~(XPC_IPI_OPENREQUEST | XPC_IPI_OPENREPLY); - - /* - * The meaningful CLOSEREQUEST connection state fields are: - * reason = reason connection is to be closed - */ - - ch->flags |= XPC_C_RCLOSEREQUEST; - - if (!(ch->flags & XPC_C_DISCONNECTING)) { - reason = args->reason; - if (reason <= xpcSuccess || reason > xpcUnknownReason) { - reason = xpcUnknownReason; - } else if (reason == xpcUnregistering) { - reason = xpcOtherUnregistering; - } - - XPC_DISCONNECT_CHANNEL(ch, reason, &irq_flags); - - DBUG_ON(IPI_flags & XPC_IPI_CLOSEREPLY); - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - xpc_process_disconnect(ch, &irq_flags); - } - - - if (IPI_flags & XPC_IPI_CLOSEREPLY) { - - dev_dbg(xpc_chan, "XPC_IPI_CLOSEREPLY received from partid=%d," - " channel=%d\n", ch->partid, ch->number); - - if (ch->flags & XPC_C_DISCONNECTED) { - DBUG_ON(part->act_state != XPC_P_DEACTIVATING); - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - DBUG_ON(!(ch->flags & XPC_C_CLOSEREQUEST)); - - if (!(ch->flags & XPC_C_RCLOSEREQUEST)) { - if ((XPC_GET_IPI_FLAGS(part->local_IPI_amo, ch_number) - & XPC_IPI_CLOSEREQUEST)) { - - DBUG_ON(ch->delayed_IPI_flags != 0); - spin_lock(&part->IPI_lock); - XPC_SET_IPI_FLAGS(part->local_IPI_amo, - ch_number, XPC_IPI_CLOSEREPLY); - spin_unlock(&part->IPI_lock); - } - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - ch->flags |= XPC_C_RCLOSEREPLY; - - if (ch->flags & XPC_C_CLOSEREPLY) { - /* both sides have finished disconnecting */ - xpc_process_disconnect(ch, &irq_flags); - } - } - - - if (IPI_flags & XPC_IPI_OPENREQUEST) { - - dev_dbg(xpc_chan, "XPC_IPI_OPENREQUEST (msg_size=%d, " - "local_nentries=%d) received from partid=%d, " - "channel=%d\n", args->msg_size, args->local_nentries, - ch->partid, ch->number); - - if (part->act_state == XPC_P_DEACTIVATING || - (ch->flags & XPC_C_ROPENREQUEST)) { - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - if (ch->flags & (XPC_C_DISCONNECTING | XPC_C_WDISCONNECT)) { - ch->delayed_IPI_flags |= XPC_IPI_OPENREQUEST; - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - DBUG_ON(!(ch->flags & (XPC_C_DISCONNECTED | - XPC_C_OPENREQUEST))); - DBUG_ON(ch->flags & (XPC_C_ROPENREQUEST | XPC_C_ROPENREPLY | - XPC_C_OPENREPLY | XPC_C_CONNECTED)); - - /* - * The meaningful OPENREQUEST connection state fields are: - * msg_size = size of channel's messages in bytes - * local_nentries = remote partition's local_nentries - */ - if (args->msg_size == 0 || args->local_nentries == 0) { - /* assume OPENREQUEST was delayed by mistake */ - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - ch->flags |= (XPC_C_ROPENREQUEST | XPC_C_CONNECTING); - ch->remote_nentries = args->local_nentries; - - - if (ch->flags & XPC_C_OPENREQUEST) { - if (args->msg_size != ch->msg_size) { - XPC_DISCONNECT_CHANNEL(ch, xpcUnequalMsgSizes, - &irq_flags); - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - } else { - ch->msg_size = args->msg_size; - - XPC_SET_REASON(ch, 0, 0); - ch->flags &= ~XPC_C_DISCONNECTED; - - atomic_inc(&part->nchannels_active); - } - - xpc_process_connect(ch, &irq_flags); - } - - - if (IPI_flags & XPC_IPI_OPENREPLY) { - - dev_dbg(xpc_chan, "XPC_IPI_OPENREPLY (local_msgqueue_pa=0x%lx, " - "local_nentries=%d, remote_nentries=%d) received from " - "partid=%d, channel=%d\n", args->local_msgqueue_pa, - args->local_nentries, args->remote_nentries, - ch->partid, ch->number); - - if (ch->flags & (XPC_C_DISCONNECTING | XPC_C_DISCONNECTED)) { - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - if (!(ch->flags & XPC_C_OPENREQUEST)) { - XPC_DISCONNECT_CHANNEL(ch, xpcOpenCloseError, - &irq_flags); - spin_unlock_irqrestore(&ch->lock, irq_flags); - return; - } - - DBUG_ON(!(ch->flags & XPC_C_ROPENREQUEST)); - DBUG_ON(ch->flags & XPC_C_CONNECTED); - - /* - * The meaningful OPENREPLY connection state fields are: - * local_msgqueue_pa = physical address of remote - * partition's local_msgqueue - * local_nentries = remote partition's local_nentries - * remote_nentries = remote partition's remote_nentries - */ - DBUG_ON(args->local_msgqueue_pa == 0); - DBUG_ON(args->local_nentries == 0); - DBUG_ON(args->remote_nentries == 0); - - ch->flags |= XPC_C_ROPENREPLY; - ch->remote_msgqueue_pa = args->local_msgqueue_pa; - - if (args->local_nentries < ch->remote_nentries) { - dev_dbg(xpc_chan, "XPC_IPI_OPENREPLY: new " - "remote_nentries=%d, old remote_nentries=%d, " - "partid=%d, channel=%d\n", - args->local_nentries, ch->remote_nentries, - ch->partid, ch->number); - - ch->remote_nentries = args->local_nentries; - } - if (args->remote_nentries < ch->local_nentries) { - dev_dbg(xpc_chan, "XPC_IPI_OPENREPLY: new " - "local_nentries=%d, old local_nentries=%d, " - "partid=%d, channel=%d\n", - args->remote_nentries, ch->local_nentries, - ch->partid, ch->number); - - ch->local_nentries = args->remote_nentries; - } - - xpc_process_connect(ch, &irq_flags); - } - - spin_unlock_irqrestore(&ch->lock, irq_flags); -} - - -/* - * Attempt to establish a channel connection to a remote partition. - */ -static enum xpc_retval -xpc_connect_channel(struct xpc_channel *ch) -{ - unsigned long irq_flags; - struct xpc_registration *registration = &xpc_registrations[ch->number]; - - - if (mutex_trylock(®istration->mutex) == 0) { - return xpcRetry; - } - - if (!XPC_CHANNEL_REGISTERED(ch->number)) { - mutex_unlock(®istration->mutex); - return xpcUnregistered; - } - - spin_lock_irqsave(&ch->lock, irq_flags); - - DBUG_ON(ch->flags & XPC_C_CONNECTED); - DBUG_ON(ch->flags & XPC_C_OPENREQUEST); - - if (ch->flags & XPC_C_DISCONNECTING) { - spin_unlock_irqrestore(&ch->lock, irq_flags); - mutex_unlock(®istration->mutex); - return ch->reason; - } - - - /* add info from the channel connect registration to the channel */ - - ch->kthreads_assigned_limit = registration->assigned_limit; - ch->kthreads_idle_limit = registration->idle_limit; - DBUG_ON(atomic_read(&ch->kthreads_assigned) != 0); - DBUG_ON(atomic_read(&ch->kthreads_idle) != 0); - DBUG_ON(atomic_read(&ch->kthreads_active) != 0); - - ch->func = registration->func; - DBUG_ON(registration->func == NULL); - ch->key = registration->key; - - ch->local_nentries = registration->nentries; - - if (ch->flags & XPC_C_ROPENREQUEST) { - if (registration->msg_size != ch->msg_size) { - /* the local and remote sides aren't the same */ - - /* - * Because XPC_DISCONNECT_CHANNEL() can block we're - * forced to up the registration sema before we unlock - * the channel lock. But that's okay here because we're - * done with the part that required the registration - * sema. XPC_DISCONNECT_CHANNEL() requires that the - * channel lock be locked and will unlock and relock - * the channel lock as needed. - */ - mutex_unlock(®istration->mutex); - XPC_DISCONNECT_CHANNEL(ch, xpcUnequalMsgSizes, - &irq_flags); - spin_unlock_irqrestore(&ch->lock, irq_flags); - return xpcUnequalMsgSizes; - } - } else { - ch->msg_size = registration->msg_size; - - XPC_SET_REASON(ch, 0, 0); - ch->flags &= ~XPC_C_DISCONNECTED; - - atomic_inc(&xpc_partitions[ch->partid].nchannels_active); - } - - mutex_unlock(®istration->mutex); - - - /* initiate the connection */ - - ch->flags |= (XPC_C_OPENREQUEST | XPC_C_CONNECTING); - xpc_IPI_send_openrequest(ch, &irq_flags); - - xpc_process_connect(ch, &irq_flags); - - spin_unlock_irqrestore(&ch->lock, irq_flags); - - return xpcSuccess; -} - - -/* - * Clear some of the msg flags in the local message queue. - */ -static inline void -xpc_clear_local_msgqueue_flags(struct xpc_channel *ch) -{ - struct xpc_msg *msg; - s64 get; - - - get = ch->w_remote_GP.get; - do { - msg = (struct xpc_msg *) ((u64) ch->local_msgqueue + - (get % ch->local_nentries) * ch->msg_size); - msg->flags = 0; - } while (++get < (volatile s64) ch->remote_GP.get); -} - - -/* - * Clear some of the msg flags in the remote message queue. - */ -static inline void -xpc_clear_remote_msgqueue_flags(struct xpc_channel *ch) -{ - struct xpc_msg *msg; - s64 put; - - - put = ch->w_remote_GP.put; - do { - msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue + - (put % ch->remote_nentries) * ch->msg_size); - msg->flags = 0; - } while (++put < (volatile s64) ch->remote_GP.put); -} - - -static void -xpc_process_msg_IPI(struct xpc_partition *part, int ch_number) -{ - struct xpc_channel *ch = &part->channels[ch_number]; - int nmsgs_sent; - - - ch->remote_GP = part->remote_GPs[ch_number]; - - - /* See what, if anything, has changed for each connected channel */ - - xpc_msgqueue_ref(ch); - - if (ch->w_remote_GP.get == ch->remote_GP.get && - ch->w_remote_GP.put == ch->remote_GP.put) { - /* nothing changed since GPs were last pulled */ - xpc_msgqueue_deref(ch); - return; - } - - if (!(ch->flags & XPC_C_CONNECTED)){ - xpc_msgqueue_deref(ch); - return; - } - - - /* - * First check to see if messages recently sent by us have been - * received by the other side. (The remote GET value will have - * changed since we last looked at it.) - */ - - if (ch->w_remote_GP.get != ch->remote_GP.get) { - - /* - * We need to notify any senders that want to be notified - * that their sent messages have been received by their - * intended recipients. We need to do this before updating - * w_remote_GP.get so that we don't allocate the same message - * queue entries prematurely (see xpc_allocate_msg()). - */ - if (atomic_read(&ch->n_to_notify) > 0) { - /* - * Notify senders that messages sent have been - * received and delivered by the other side. - */ - xpc_notify_senders(ch, xpcMsgDelivered, - ch->remote_GP.get); - } - - /* - * Clear msg->flags in previously sent messages, so that - * they're ready for xpc_allocate_msg(). - */ - xpc_clear_local_msgqueue_flags(ch); - - ch->w_remote_GP.get = ch->remote_GP.get; - - dev_dbg(xpc_chan, "w_remote_GP.get changed to %ld, partid=%d, " - "channel=%d\n", ch->w_remote_GP.get, ch->partid, - ch->number); - - /* - * If anyone was waiting for message queue entries to become - * available, wake them up. - */ - if (atomic_read(&ch->n_on_msg_allocate_wq) > 0) { - wake_up(&ch->msg_allocate_wq); - } - } - - - /* - * Now check for newly sent messages by the other side. (The remote - * PUT value will have changed since we last looked at it.) - */ - - if (ch->w_remote_GP.put != ch->remote_GP.put) { - /* - * Clear msg->flags in previously received messages, so that - * they're ready for xpc_get_deliverable_msg(). - */ - xpc_clear_remote_msgqueue_flags(ch); - - ch->w_remote_GP.put = ch->remote_GP.put; - - dev_dbg(xpc_chan, "w_remote_GP.put changed to %ld, partid=%d, " - "channel=%d\n", ch->w_remote_GP.put, ch->partid, - ch->number); - - nmsgs_sent = ch->w_remote_GP.put - ch->w_local_GP.get; - if (nmsgs_sent > 0) { - dev_dbg(xpc_chan, "msgs waiting to be copied and " - "delivered=%d, partid=%d, channel=%d\n", - nmsgs_sent, ch->partid, ch->number); - - if (ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) { - xpc_activate_kthreads(ch, nmsgs_sent); - } - } - } - - xpc_msgqueue_deref(ch); -} - - -void -xpc_process_channel_activity(struct xpc_partition *part) -{ - unsigned long irq_flags; - u64 IPI_amo, IPI_flags; - struct xpc_channel *ch; - int ch_number; - u32 ch_flags; - - - IPI_amo = xpc_get_IPI_flags(part); - - /* - * Initiate channel connections for registered channels. - * - * For each connected channel that has pending messages activate idle - * kthreads and/or create new kthreads as needed. - */ - - for (ch_number = 0; ch_number < part->nchannels; ch_number++) { - ch = &part->channels[ch_number]; - - - /* - * Process any open or close related IPI flags, and then deal - * with connecting or disconnecting the channel as required. - */ - - IPI_flags = XPC_GET_IPI_FLAGS(IPI_amo, ch_number); - - if (XPC_ANY_OPENCLOSE_IPI_FLAGS_SET(IPI_flags)) { - xpc_process_openclose_IPI(part, ch_number, IPI_flags); - } - - ch_flags = ch->flags; /* need an atomic snapshot of flags */ - - if (ch_flags & XPC_C_DISCONNECTING) { - spin_lock_irqsave(&ch->lock, irq_flags); - xpc_process_disconnect(ch, &irq_flags); - spin_unlock_irqrestore(&ch->lock, irq_flags); - continue; - } - - if (part->act_state == XPC_P_DEACTIVATING) { - continue; - } - - if (!(ch_flags & XPC_C_CONNECTED)) { - if (!(ch_flags & XPC_C_OPENREQUEST)) { - DBUG_ON(ch_flags & XPC_C_SETUP); - (void) xpc_connect_channel(ch); - } else { - spin_lock_irqsave(&ch->lock, irq_flags); - xpc_process_connect(ch, &irq_flags); - spin_unlock_irqrestore(&ch->lock, irq_flags); - } - continue; - } - - - /* - * Process any message related IPI flags, this may involve the - * activation of kthreads to deliver any pending messages sent - * from the other partition. - */ - - if (XPC_ANY_MSG_IPI_FLAGS_SET(IPI_flags)) { - xpc_process_msg_IPI(part, ch_number); - } - } -} - - -/* - * XPC's heartbeat code calls this function to inform XPC that a partition is - * going down. XPC responds by tearing down the XPartition Communication - * infrastructure used for the just downed partition. - * - * XPC's heartbeat code will never call this function and xpc_partition_up() - * at the same time. Nor will it ever make multiple calls to either function - * at the same time. - */ -void -xpc_partition_going_down(struct xpc_partition *part, enum xpc_retval reason) -{ - unsigned long irq_flags; - int ch_number; - struct xpc_channel *ch; - - - dev_dbg(xpc_chan, "deactivating partition %d, reason=%d\n", - XPC_PARTID(part), reason); - - if (!xpc_part_ref(part)) { - /* infrastructure for this partition isn't currently set up */ - return; - } - - - /* disconnect channels associated with the partition going down */ - - for (ch_number = 0; ch_number < part->nchannels; ch_number++) { - ch = &part->channels[ch_number]; - - xpc_msgqueue_ref(ch); - spin_lock_irqsave(&ch->lock, irq_flags); - - XPC_DISCONNECT_CHANNEL(ch, reason, &irq_flags); - - spin_unlock_irqrestore(&ch->lock, irq_flags); - xpc_msgqueue_deref(ch); - } - - xpc_wakeup_channel_mgr(part); - - xpc_part_deref(part); -} - - -/* - * Teardown the infrastructure necessary to support XPartition Communication - * between the specified remote partition and the local one. - */ -void -xpc_teardown_infrastructure(struct xpc_partition *part) -{ - partid_t partid = XPC_PARTID(part); - - - /* - * We start off by making this partition inaccessible to local - * processes by marking it as no longer setup. Then we make it - * inaccessible to remote processes by clearing the XPC per partition - * specific variable's magic # (which indicates that these variables - * are no longer valid) and by ignoring all XPC notify IPIs sent to - * this partition. - */ - - DBUG_ON(atomic_read(&part->nchannels_engaged) != 0); - DBUG_ON(atomic_read(&part->nchannels_active) != 0); - DBUG_ON(part->setup_state != XPC_P_SETUP); - part->setup_state = XPC_P_WTEARDOWN; - - xpc_vars_part[partid].magic = 0; - - - free_irq(SGI_XPC_NOTIFY, (void *) (u64) partid); - - - /* - * Before proceding with the teardown we have to wait until all - * existing references cease. - */ - wait_event(part->teardown_wq, (atomic_read(&part->references) == 0)); - - - /* now we can begin tearing down the infrastructure */ - - part->setup_state = XPC_P_TORNDOWN; - - /* in case we've still got outstanding timers registered... */ - del_timer_sync(&part->dropped_IPI_timer); - - kfree(part->remote_openclose_args_base); - part->remote_openclose_args = NULL; - kfree(part->local_openclose_args_base); - part->local_openclose_args = NULL; - kfree(part->remote_GPs_base); - part->remote_GPs = NULL; - kfree(part->local_GPs_base); - part->local_GPs = NULL; - kfree(part->channels); - part->channels = NULL; - part->local_IPI_amo_va = NULL; -} - - -/* - * Called by XP at the time of channel connection registration to cause - * XPC to establish connections to all currently active partitions. - */ -void -xpc_initiate_connect(int ch_number) -{ - partid_t partid; - struct xpc_partition *part; - struct xpc_channel *ch; - - - DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS); - - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - part = &xpc_partitions[partid]; - - if (xpc_part_ref(part)) { - ch = &part->channels[ch_number]; - - /* - * Initiate the establishment of a connection on the - * newly registered channel to the remote partition. - */ - xpc_wakeup_channel_mgr(part); - xpc_part_deref(part); - } - } -} - - -void -xpc_connected_callout(struct xpc_channel *ch) -{ - /* let the registerer know that a connection has been established */ - - if (ch->func != NULL) { - dev_dbg(xpc_chan, "ch->func() called, reason=xpcConnected, " - "partid=%d, channel=%d\n", ch->partid, ch->number); - - ch->func(xpcConnected, ch->partid, ch->number, - (void *) (u64) ch->local_nentries, ch->key); - - dev_dbg(xpc_chan, "ch->func() returned, reason=xpcConnected, " - "partid=%d, channel=%d\n", ch->partid, ch->number); - } -} - - -/* - * Called by XP at the time of channel connection unregistration to cause - * XPC to teardown all current connections for the specified channel. - * - * Before returning xpc_initiate_disconnect() will wait until all connections - * on the specified channel have been closed/torndown. So the caller can be - * assured that they will not be receiving any more callouts from XPC to the - * function they registered via xpc_connect(). - * - * Arguments: - * - * ch_number - channel # to unregister. - */ -void -xpc_initiate_disconnect(int ch_number) -{ - unsigned long irq_flags; - partid_t partid; - struct xpc_partition *part; - struct xpc_channel *ch; - - - DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS); - - /* initiate the channel disconnect for every active partition */ - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - part = &xpc_partitions[partid]; - - if (xpc_part_ref(part)) { - ch = &part->channels[ch_number]; - xpc_msgqueue_ref(ch); - - spin_lock_irqsave(&ch->lock, irq_flags); - - if (!(ch->flags & XPC_C_DISCONNECTED)) { - ch->flags |= XPC_C_WDISCONNECT; - - XPC_DISCONNECT_CHANNEL(ch, xpcUnregistering, - &irq_flags); - } - - spin_unlock_irqrestore(&ch->lock, irq_flags); - - xpc_msgqueue_deref(ch); - xpc_part_deref(part); - } - } - - xpc_disconnect_wait(ch_number); -} - - -/* - * To disconnect a channel, and reflect it back to all who may be waiting. - * - * An OPEN is not allowed until XPC_C_DISCONNECTING is cleared by - * xpc_process_disconnect(), and if set, XPC_C_WDISCONNECT is cleared by - * xpc_disconnect_wait(). - * - * THE CHANNEL IS TO BE LOCKED BY THE CALLER AND WILL REMAIN LOCKED UPON RETURN. - */ -void -xpc_disconnect_channel(const int line, struct xpc_channel *ch, - enum xpc_retval reason, unsigned long *irq_flags) -{ - u32 channel_was_connected = (ch->flags & XPC_C_CONNECTED); - - - DBUG_ON(!spin_is_locked(&ch->lock)); - - if (ch->flags & (XPC_C_DISCONNECTING | XPC_C_DISCONNECTED)) { - return; - } - DBUG_ON(!(ch->flags & (XPC_C_CONNECTING | XPC_C_CONNECTED))); - - dev_dbg(xpc_chan, "reason=%d, line=%d, partid=%d, channel=%d\n", - reason, line, ch->partid, ch->number); - - XPC_SET_REASON(ch, reason, line); - - ch->flags |= (XPC_C_CLOSEREQUEST | XPC_C_DISCONNECTING); - /* some of these may not have been set */ - ch->flags &= ~(XPC_C_OPENREQUEST | XPC_C_OPENREPLY | - XPC_C_ROPENREQUEST | XPC_C_ROPENREPLY | - XPC_C_CONNECTING | XPC_C_CONNECTED); - - xpc_IPI_send_closerequest(ch, irq_flags); - - if (channel_was_connected) { - ch->flags |= XPC_C_WASCONNECTED; - } - - spin_unlock_irqrestore(&ch->lock, *irq_flags); - - /* wake all idle kthreads so they can exit */ - if (atomic_read(&ch->kthreads_idle) > 0) { - wake_up_all(&ch->idle_wq); - } - - /* wake those waiting to allocate an entry from the local msg queue */ - if (atomic_read(&ch->n_on_msg_allocate_wq) > 0) { - wake_up(&ch->msg_allocate_wq); - } - - spin_lock_irqsave(&ch->lock, *irq_flags); -} - - -void -xpc_disconnect_callout(struct xpc_channel *ch, enum xpc_retval reason) -{ - /* - * Let the channel's registerer know that the channel is being - * disconnected. We don't want to do this if the registerer was never - * informed of a connection being made. - */ - - if (ch->func != NULL) { - dev_dbg(xpc_chan, "ch->func() called, reason=%d, partid=%d, " - "channel=%d\n", reason, ch->partid, ch->number); - - ch->func(reason, ch->partid, ch->number, NULL, ch->key); - - dev_dbg(xpc_chan, "ch->func() returned, reason=%d, partid=%d, " - "channel=%d\n", reason, ch->partid, ch->number); - } -} - - -/* - * Wait for a message entry to become available for the specified channel, - * but don't wait any longer than 1 jiffy. - */ -static enum xpc_retval -xpc_allocate_msg_wait(struct xpc_channel *ch) -{ - enum xpc_retval ret; - - - if (ch->flags & XPC_C_DISCONNECTING) { - DBUG_ON(ch->reason == xpcInterrupted); // >>> Is this true? - return ch->reason; - } - - atomic_inc(&ch->n_on_msg_allocate_wq); - ret = interruptible_sleep_on_timeout(&ch->msg_allocate_wq, 1); - atomic_dec(&ch->n_on_msg_allocate_wq); - - if (ch->flags & XPC_C_DISCONNECTING) { - ret = ch->reason; - DBUG_ON(ch->reason == xpcInterrupted); // >>> Is this true? - } else if (ret == 0) { - ret = xpcTimeout; - } else { - ret = xpcInterrupted; - } - - return ret; -} - - -/* - * Allocate an entry for a message from the message queue associated with the - * specified channel. - */ -static enum xpc_retval -xpc_allocate_msg(struct xpc_channel *ch, u32 flags, - struct xpc_msg **address_of_msg) -{ - struct xpc_msg *msg; - enum xpc_retval ret; - s64 put; - - - /* this reference will be dropped in xpc_send_msg() */ - xpc_msgqueue_ref(ch); - - if (ch->flags & XPC_C_DISCONNECTING) { - xpc_msgqueue_deref(ch); - return ch->reason; - } - if (!(ch->flags & XPC_C_CONNECTED)) { - xpc_msgqueue_deref(ch); - return xpcNotConnected; - } - - - /* - * Get the next available message entry from the local message queue. - * If none are available, we'll make sure that we grab the latest - * GP values. - */ - ret = xpcTimeout; - - while (1) { - - put = (volatile s64) ch->w_local_GP.put; - if (put - (volatile s64) ch->w_remote_GP.get < - ch->local_nentries) { - - /* There are available message entries. We need to try - * to secure one for ourselves. We'll do this by trying - * to increment w_local_GP.put as long as someone else - * doesn't beat us to it. If they do, we'll have to - * try again. - */ - if (cmpxchg(&ch->w_local_GP.put, put, put + 1) == - put) { - /* we got the entry referenced by put */ - break; - } - continue; /* try again */ - } - - - /* - * There aren't any available msg entries at this time. - * - * In waiting for a message entry to become available, - * we set a timeout in case the other side is not - * sending completion IPIs. This lets us fake an IPI - * that will cause the IPI handler to fetch the latest - * GP values as if an IPI was sent by the other side. - */ - if (ret == xpcTimeout) { - xpc_IPI_send_local_msgrequest(ch); - } - - if (flags & XPC_NOWAIT) { - xpc_msgqueue_deref(ch); - return xpcNoWait; - } - - ret = xpc_allocate_msg_wait(ch); - if (ret != xpcInterrupted && ret != xpcTimeout) { - xpc_msgqueue_deref(ch); - return ret; - } - } - - - /* get the message's address and initialize it */ - msg = (struct xpc_msg *) ((u64) ch->local_msgqueue + - (put % ch->local_nentries) * ch->msg_size); - - - DBUG_ON(msg->flags != 0); - msg->number = put; - - dev_dbg(xpc_chan, "w_local_GP.put changed to %ld; msg=0x%p, " - "msg_number=%ld, partid=%d, channel=%d\n", put + 1, - (void *) msg, msg->number, ch->partid, ch->number); - - *address_of_msg = msg; - - return xpcSuccess; -} - - -/* - * Allocate an entry for a message from the message queue associated with the - * specified channel. NOTE that this routine can sleep waiting for a message - * entry to become available. To not sleep, pass in the XPC_NOWAIT flag. - * - * Arguments: - * - * partid - ID of partition to which the channel is connected. - * ch_number - channel #. - * flags - see xpc.h for valid flags. - * payload - address of the allocated payload area pointer (filled in on - * return) in which the user-defined message is constructed. - */ -enum xpc_retval -xpc_initiate_allocate(partid_t partid, int ch_number, u32 flags, void **payload) -{ - struct xpc_partition *part = &xpc_partitions[partid]; - enum xpc_retval ret = xpcUnknownReason; - struct xpc_msg *msg = NULL; - - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - DBUG_ON(ch_number < 0 || ch_number >= part->nchannels); - - *payload = NULL; - - if (xpc_part_ref(part)) { - ret = xpc_allocate_msg(&part->channels[ch_number], flags, &msg); - xpc_part_deref(part); - - if (msg != NULL) { - *payload = &msg->payload; - } - } - - return ret; -} - - -/* - * Now we actually send the messages that are ready to be sent by advancing - * the local message queue's Put value and then send an IPI to the recipient - * partition. - */ -static void -xpc_send_msgs(struct xpc_channel *ch, s64 initial_put) -{ - struct xpc_msg *msg; - s64 put = initial_put + 1; - int send_IPI = 0; - - - while (1) { - - while (1) { - if (put == (volatile s64) ch->w_local_GP.put) { - break; - } - - msg = (struct xpc_msg *) ((u64) ch->local_msgqueue + - (put % ch->local_nentries) * ch->msg_size); - - if (!(msg->flags & XPC_M_READY)) { - break; - } - - put++; - } - - if (put == initial_put) { - /* nothing's changed */ - break; - } - - if (cmpxchg_rel(&ch->local_GP->put, initial_put, put) != - initial_put) { - /* someone else beat us to it */ - DBUG_ON((volatile s64) ch->local_GP->put < initial_put); - break; - } - - /* we just set the new value of local_GP->put */ - - dev_dbg(xpc_chan, "local_GP->put changed to %ld, partid=%d, " - "channel=%d\n", put, ch->partid, ch->number); - - send_IPI = 1; - - /* - * We need to ensure that the message referenced by - * local_GP->put is not XPC_M_READY or that local_GP->put - * equals w_local_GP.put, so we'll go have a look. - */ - initial_put = put; - } - - if (send_IPI) { - xpc_IPI_send_msgrequest(ch); - } -} - - -/* - * Common code that does the actual sending of the message by advancing the - * local message queue's Put value and sends an IPI to the partition the - * message is being sent to. - */ -static enum xpc_retval -xpc_send_msg(struct xpc_channel *ch, struct xpc_msg *msg, u8 notify_type, - xpc_notify_func func, void *key) -{ - enum xpc_retval ret = xpcSuccess; - struct xpc_notify *notify = notify; - s64 put, msg_number = msg->number; - - - DBUG_ON(notify_type == XPC_N_CALL && func == NULL); - DBUG_ON((((u64) msg - (u64) ch->local_msgqueue) / ch->msg_size) != - msg_number % ch->local_nentries); - DBUG_ON(msg->flags & XPC_M_READY); - - if (ch->flags & XPC_C_DISCONNECTING) { - /* drop the reference grabbed in xpc_allocate_msg() */ - xpc_msgqueue_deref(ch); - return ch->reason; - } - - if (notify_type != 0) { - /* - * Tell the remote side to send an ACK interrupt when the - * message has been delivered. - */ - msg->flags |= XPC_M_INTERRUPT; - - atomic_inc(&ch->n_to_notify); - - notify = &ch->notify_queue[msg_number % ch->local_nentries]; - notify->func = func; - notify->key = key; - notify->type = notify_type; - - // >>> is a mb() needed here? - - if (ch->flags & XPC_C_DISCONNECTING) { - /* - * An error occurred between our last error check and - * this one. We will try to clear the type field from - * the notify entry. If we succeed then - * xpc_disconnect_channel() didn't already process - * the notify entry. - */ - if (cmpxchg(¬ify->type, notify_type, 0) == - notify_type) { - atomic_dec(&ch->n_to_notify); - ret = ch->reason; - } - - /* drop the reference grabbed in xpc_allocate_msg() */ - xpc_msgqueue_deref(ch); - return ret; - } - } - - msg->flags |= XPC_M_READY; - - /* - * The preceding store of msg->flags must occur before the following - * load of ch->local_GP->put. - */ - mb(); - - /* see if the message is next in line to be sent, if so send it */ - - put = ch->local_GP->put; - if (put == msg_number) { - xpc_send_msgs(ch, put); - } - - /* drop the reference grabbed in xpc_allocate_msg() */ - xpc_msgqueue_deref(ch); - return ret; -} - - -/* - * Send a message previously allocated using xpc_initiate_allocate() on the - * specified channel connected to the specified partition. - * - * This routine will not wait for the message to be received, nor will - * notification be given when it does happen. Once this routine has returned - * the message entry allocated via xpc_initiate_allocate() is no longer - * accessable to the caller. - * - * This routine, although called by users, does not call xpc_part_ref() to - * ensure that the partition infrastructure is in place. It relies on the - * fact that we called xpc_msgqueue_ref() in xpc_allocate_msg(). - * - * Arguments: - * - * partid - ID of partition to which the channel is connected. - * ch_number - channel # to send message on. - * payload - pointer to the payload area allocated via - * xpc_initiate_allocate(). - */ -enum xpc_retval -xpc_initiate_send(partid_t partid, int ch_number, void *payload) -{ - struct xpc_partition *part = &xpc_partitions[partid]; - struct xpc_msg *msg = XPC_MSG_ADDRESS(payload); - enum xpc_retval ret; - - - dev_dbg(xpc_chan, "msg=0x%p, partid=%d, channel=%d\n", (void *) msg, - partid, ch_number); - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - DBUG_ON(ch_number < 0 || ch_number >= part->nchannels); - DBUG_ON(msg == NULL); - - ret = xpc_send_msg(&part->channels[ch_number], msg, 0, NULL, NULL); - - return ret; -} - - -/* - * Send a message previously allocated using xpc_initiate_allocate on the - * specified channel connected to the specified partition. - * - * This routine will not wait for the message to be sent. Once this routine - * has returned the message entry allocated via xpc_initiate_allocate() is no - * longer accessable to the caller. - * - * Once the remote end of the channel has received the message, the function - * passed as an argument to xpc_initiate_send_notify() will be called. This - * allows the sender to free up or re-use any buffers referenced by the - * message, but does NOT mean the message has been processed at the remote - * end by a receiver. - * - * If this routine returns an error, the caller's function will NOT be called. - * - * This routine, although called by users, does not call xpc_part_ref() to - * ensure that the partition infrastructure is in place. It relies on the - * fact that we called xpc_msgqueue_ref() in xpc_allocate_msg(). - * - * Arguments: - * - * partid - ID of partition to which the channel is connected. - * ch_number - channel # to send message on. - * payload - pointer to the payload area allocated via - * xpc_initiate_allocate(). - * func - function to call with asynchronous notification of message - * receipt. THIS FUNCTION MUST BE NON-BLOCKING. - * key - user-defined key to be passed to the function when it's called. - */ -enum xpc_retval -xpc_initiate_send_notify(partid_t partid, int ch_number, void *payload, - xpc_notify_func func, void *key) -{ - struct xpc_partition *part = &xpc_partitions[partid]; - struct xpc_msg *msg = XPC_MSG_ADDRESS(payload); - enum xpc_retval ret; - - - dev_dbg(xpc_chan, "msg=0x%p, partid=%d, channel=%d\n", (void *) msg, - partid, ch_number); - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - DBUG_ON(ch_number < 0 || ch_number >= part->nchannels); - DBUG_ON(msg == NULL); - DBUG_ON(func == NULL); - - ret = xpc_send_msg(&part->channels[ch_number], msg, XPC_N_CALL, - func, key); - return ret; -} - - -static struct xpc_msg * -xpc_pull_remote_msg(struct xpc_channel *ch, s64 get) -{ - struct xpc_partition *part = &xpc_partitions[ch->partid]; - struct xpc_msg *remote_msg, *msg; - u32 msg_index, nmsgs; - u64 msg_offset; - enum xpc_retval ret; - - - if (mutex_lock_interruptible(&ch->msg_to_pull_mutex) != 0) { - /* we were interrupted by a signal */ - return NULL; - } - - while (get >= ch->next_msg_to_pull) { - - /* pull as many messages as are ready and able to be pulled */ - - msg_index = ch->next_msg_to_pull % ch->remote_nentries; - - DBUG_ON(ch->next_msg_to_pull >= - (volatile s64) ch->w_remote_GP.put); - nmsgs = (volatile s64) ch->w_remote_GP.put - - ch->next_msg_to_pull; - if (msg_index + nmsgs > ch->remote_nentries) { - /* ignore the ones that wrap the msg queue for now */ - nmsgs = ch->remote_nentries - msg_index; - } - - msg_offset = msg_index * ch->msg_size; - msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue + - msg_offset); - remote_msg = (struct xpc_msg *) (ch->remote_msgqueue_pa + - msg_offset); - - if ((ret = xpc_pull_remote_cachelines(part, msg, remote_msg, - nmsgs * ch->msg_size)) != xpcSuccess) { - - dev_dbg(xpc_chan, "failed to pull %d msgs starting with" - " msg %ld from partition %d, channel=%d, " - "ret=%d\n", nmsgs, ch->next_msg_to_pull, - ch->partid, ch->number, ret); - - XPC_DEACTIVATE_PARTITION(part, ret); - - mutex_unlock(&ch->msg_to_pull_mutex); - return NULL; - } - - mb(); /* >>> this may not be needed, we're not sure */ - - ch->next_msg_to_pull += nmsgs; - } - - mutex_unlock(&ch->msg_to_pull_mutex); - - /* return the message we were looking for */ - msg_offset = (get % ch->remote_nentries) * ch->msg_size; - msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue + msg_offset); - - return msg; -} - - -/* - * Get a message to be delivered. - */ -static struct xpc_msg * -xpc_get_deliverable_msg(struct xpc_channel *ch) -{ - struct xpc_msg *msg = NULL; - s64 get; - - - do { - if ((volatile u32) ch->flags & XPC_C_DISCONNECTING) { - break; - } - - get = (volatile s64) ch->w_local_GP.get; - if (get == (volatile s64) ch->w_remote_GP.put) { - break; - } - - /* There are messages waiting to be pulled and delivered. - * We need to try to secure one for ourselves. We'll do this - * by trying to increment w_local_GP.get and hope that no one - * else beats us to it. If they do, we'll we'll simply have - * to try again for the next one. - */ - - if (cmpxchg(&ch->w_local_GP.get, get, get + 1) == get) { - /* we got the entry referenced by get */ - - dev_dbg(xpc_chan, "w_local_GP.get changed to %ld, " - "partid=%d, channel=%d\n", get + 1, - ch->partid, ch->number); - - /* pull the message from the remote partition */ - - msg = xpc_pull_remote_msg(ch, get); - - DBUG_ON(msg != NULL && msg->number != get); - DBUG_ON(msg != NULL && (msg->flags & XPC_M_DONE)); - DBUG_ON(msg != NULL && !(msg->flags & XPC_M_READY)); - - break; - } - - } while (1); - - return msg; -} - - -/* - * Deliver a message to its intended recipient. - */ -void -xpc_deliver_msg(struct xpc_channel *ch) -{ - struct xpc_msg *msg; - - - if ((msg = xpc_get_deliverable_msg(ch)) != NULL) { - - /* - * This ref is taken to protect the payload itself from being - * freed before the user is finished with it, which the user - * indicates by calling xpc_initiate_received(). - */ - xpc_msgqueue_ref(ch); - - atomic_inc(&ch->kthreads_active); - - if (ch->func != NULL) { - dev_dbg(xpc_chan, "ch->func() called, msg=0x%p, " - "msg_number=%ld, partid=%d, channel=%d\n", - (void *) msg, msg->number, ch->partid, - ch->number); - - /* deliver the message to its intended recipient */ - ch->func(xpcMsgReceived, ch->partid, ch->number, - &msg->payload, ch->key); - - dev_dbg(xpc_chan, "ch->func() returned, msg=0x%p, " - "msg_number=%ld, partid=%d, channel=%d\n", - (void *) msg, msg->number, ch->partid, - ch->number); - } - - atomic_dec(&ch->kthreads_active); - } -} - - -/* - * Now we actually acknowledge the messages that have been delivered and ack'd - * by advancing the cached remote message queue's Get value and if requested - * send an IPI to the message sender's partition. - */ -static void -xpc_acknowledge_msgs(struct xpc_channel *ch, s64 initial_get, u8 msg_flags) -{ - struct xpc_msg *msg; - s64 get = initial_get + 1; - int send_IPI = 0; - - - while (1) { - - while (1) { - if (get == (volatile s64) ch->w_local_GP.get) { - break; - } - - msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue + - (get % ch->remote_nentries) * ch->msg_size); - - if (!(msg->flags & XPC_M_DONE)) { - break; - } - - msg_flags |= msg->flags; - get++; - } - - if (get == initial_get) { - /* nothing's changed */ - break; - } - - if (cmpxchg_rel(&ch->local_GP->get, initial_get, get) != - initial_get) { - /* someone else beat us to it */ - DBUG_ON((volatile s64) ch->local_GP->get <= - initial_get); - break; - } - - /* we just set the new value of local_GP->get */ - - dev_dbg(xpc_chan, "local_GP->get changed to %ld, partid=%d, " - "channel=%d\n", get, ch->partid, ch->number); - - send_IPI = (msg_flags & XPC_M_INTERRUPT); - - /* - * We need to ensure that the message referenced by - * local_GP->get is not XPC_M_DONE or that local_GP->get - * equals w_local_GP.get, so we'll go have a look. - */ - initial_get = get; - } - - if (send_IPI) { - xpc_IPI_send_msgrequest(ch); - } -} - - -/* - * Acknowledge receipt of a delivered message. - * - * If a message has XPC_M_INTERRUPT set, send an interrupt to the partition - * that sent the message. - * - * This function, although called by users, does not call xpc_part_ref() to - * ensure that the partition infrastructure is in place. It relies on the - * fact that we called xpc_msgqueue_ref() in xpc_deliver_msg(). - * - * Arguments: - * - * partid - ID of partition to which the channel is connected. - * ch_number - channel # message received on. - * payload - pointer to the payload area allocated via - * xpc_initiate_allocate(). - */ -void -xpc_initiate_received(partid_t partid, int ch_number, void *payload) -{ - struct xpc_partition *part = &xpc_partitions[partid]; - struct xpc_channel *ch; - struct xpc_msg *msg = XPC_MSG_ADDRESS(payload); - s64 get, msg_number = msg->number; - - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - DBUG_ON(ch_number < 0 || ch_number >= part->nchannels); - - ch = &part->channels[ch_number]; - - dev_dbg(xpc_chan, "msg=0x%p, msg_number=%ld, partid=%d, channel=%d\n", - (void *) msg, msg_number, ch->partid, ch->number); - - DBUG_ON((((u64) msg - (u64) ch->remote_msgqueue) / ch->msg_size) != - msg_number % ch->remote_nentries); - DBUG_ON(msg->flags & XPC_M_DONE); - - msg->flags |= XPC_M_DONE; - - /* - * The preceding store of msg->flags must occur before the following - * load of ch->local_GP->get. - */ - mb(); - - /* - * See if this message is next in line to be acknowledged as having - * been delivered. - */ - get = ch->local_GP->get; - if (get == msg_number) { - xpc_acknowledge_msgs(ch, get, msg->flags); - } - - /* the call to xpc_msgqueue_ref() was done by xpc_deliver_msg() */ - xpc_msgqueue_deref(ch); -} - diff --git a/arch/ia64/sn/kernel/xpc_main.c b/arch/ia64/sn/kernel/xpc_main.c deleted file mode 100644 index 4d026f9dd98..00000000000 --- a/arch/ia64/sn/kernel/xpc_main.c +++ /dev/null @@ -1,1428 +0,0 @@ -/* - * This file is subject to the terms and conditions of the GNU General Public - * License. See the file "COPYING" in the main directory of this archive - * for more details. - * - * Copyright (c) 2004-2006 Silicon Graphics, Inc. All Rights Reserved. - */ - - -/* - * Cross Partition Communication (XPC) support - standard version. - * - * XPC provides a message passing capability that crosses partition - * boundaries. This module is made up of two parts: - * - * partition This part detects the presence/absence of other - * partitions. It provides a heartbeat and monitors - * the heartbeats of other partitions. - * - * channel This part manages the channels and sends/receives - * messages across them to/from other partitions. - * - * There are a couple of additional functions residing in XP, which - * provide an interface to XPC for its users. - * - * - * Caveats: - * - * . We currently have no way to determine which nasid an IPI came - * from. Thus, xpc_IPI_send() does a remote AMO write followed by - * an IPI. The AMO indicates where data is to be pulled from, so - * after the IPI arrives, the remote partition checks the AMO word. - * The IPI can actually arrive before the AMO however, so other code - * must periodically check for this case. Also, remote AMO operations - * do not reliably time out. Thus we do a remote PIO read solely to - * know whether the remote partition is down and whether we should - * stop sending IPIs to it. This remote PIO read operation is set up - * in a special nofault region so SAL knows to ignore (and cleanup) - * any errors due to the remote AMO write, PIO read, and/or PIO - * write operations. - * - * If/when new hardware solves this IPI problem, we should abandon - * the current approach. - * - */ - - -#include <linux/kernel.h> -#include <linux/module.h> -#include <linux/init.h> -#include <linux/sched.h> -#include <linux/syscalls.h> -#include <linux/cache.h> -#include <linux/interrupt.h> -#include <linux/delay.h> -#include <linux/reboot.h> -#include <linux/completion.h> -#include <asm/sn/intr.h> -#include <asm/sn/sn_sal.h> -#include <asm/kdebug.h> -#include <asm/uaccess.h> -#include <asm/sn/xpc.h> - - -/* define two XPC debug device structures to be used with dev_dbg() et al */ - -struct device_driver xpc_dbg_name = { - .name = "xpc" -}; - -struct device xpc_part_dbg_subname = { - .bus_id = {0}, /* set to "part" at xpc_init() time */ - .driver = &xpc_dbg_name -}; - -struct device xpc_chan_dbg_subname = { - .bus_id = {0}, /* set to "chan" at xpc_init() time */ - .driver = &xpc_dbg_name -}; - -struct device *xpc_part = &xpc_part_dbg_subname; -struct device *xpc_chan = &xpc_chan_dbg_subname; - - -static int xpc_kdebug_ignore; - - -/* systune related variables for /proc/sys directories */ - -static int xpc_hb_interval = XPC_HB_DEFAULT_INTERVAL; -static int xpc_hb_min_interval = 1; -static int xpc_hb_max_interval = 10; - -static int xpc_hb_check_interval = XPC_HB_CHECK_DEFAULT_INTERVAL; -static int xpc_hb_check_min_interval = 10; -static int xpc_hb_check_max_interval = 120; - -int xpc_disengage_request_timelimit = XPC_DISENGAGE_REQUEST_DEFAULT_TIMELIMIT; -static int xpc_disengage_request_min_timelimit = 0; -static int xpc_disengage_request_max_timelimit = 120; - -static ctl_table xpc_sys_xpc_hb_dir[] = { - { - 1, - "hb_interval", - &xpc_hb_interval, - sizeof(int), - 0644, - NULL, - &proc_dointvec_minmax, - &sysctl_intvec, - NULL, - &xpc_hb_min_interval, - &xpc_hb_max_interval - }, - { - 2, - "hb_check_interval", - &xpc_hb_check_interval, - sizeof(int), - 0644, - NULL, - &proc_dointvec_minmax, - &sysctl_intvec, - NULL, - &xpc_hb_check_min_interval, - &xpc_hb_check_max_interval - }, - {0} -}; -static ctl_table xpc_sys_xpc_dir[] = { - { - 1, - "hb", - NULL, - 0, - 0555, - xpc_sys_xpc_hb_dir - }, - { - 2, - "disengage_request_timelimit", - &xpc_disengage_request_timelimit, - sizeof(int), - 0644, - NULL, - &proc_dointvec_minmax, - &sysctl_intvec, - NULL, - &xpc_disengage_request_min_timelimit, - &xpc_disengage_request_max_timelimit - }, - {0} -}; -static ctl_table xpc_sys_dir[] = { - { - 1, - "xpc", - NULL, - 0, - 0555, - xpc_sys_xpc_dir - }, - {0} -}; -static struct ctl_table_header *xpc_sysctl; - -/* non-zero if any remote partition disengage request was timed out */ -int xpc_disengage_request_timedout; - -/* #of IRQs received */ -static atomic_t xpc_act_IRQ_rcvd; - -/* IRQ handler notifies this wait queue on receipt of an IRQ */ -static DECLARE_WAIT_QUEUE_HEAD(xpc_act_IRQ_wq); - -static unsigned long xpc_hb_check_timeout; - -/* notification that the xpc_hb_checker thread has exited */ -static DECLARE_COMPLETION(xpc_hb_checker_exited); - -/* notification that the xpc_discovery thread has exited */ -static DECLARE_COMPLETION(xpc_discovery_exited); - - -static struct timer_list xpc_hb_timer; - - -static void xpc_kthread_waitmsgs(struct xpc_partition *, struct xpc_channel *); - - -static int xpc_system_reboot(struct notifier_block *, unsigned long, void *); -static struct notifier_block xpc_reboot_notifier = { - .notifier_call = xpc_system_reboot, -}; - -static int xpc_system_die(struct notifier_block *, unsigned long, void *); -static struct notifier_block xpc_die_notifier = { - .notifier_call = xpc_system_die, -}; - - -/* - * Timer function to enforce the timelimit on the partition disengage request. - */ -static void -xpc_timeout_partition_disengage_request(unsigned long data) -{ - struct xpc_partition *part = (struct xpc_partition *) data; - - - DBUG_ON(jiffies < part->disengage_request_timeout); - - (void) xpc_partition_disengaged(part); - - DBUG_ON(part->disengage_request_timeout != 0); - DBUG_ON(xpc_partition_engaged(1UL << XPC_PARTID(part)) != 0); -} - - -/* - * Notify the heartbeat check thread that an IRQ has been received. - */ -static irqreturn_t -xpc_act_IRQ_handler(int irq, void *dev_id, struct pt_regs *regs) -{ - atomic_inc(&xpc_act_IRQ_rcvd); - wake_up_interruptible(&xpc_act_IRQ_wq); - return IRQ_HANDLED; -} - - -/* - * Timer to produce the heartbeat. The timer structures function is - * already set when this is initially called. A tunable is used to - * specify when the next timeout should occur. - */ -static void -xpc_hb_beater(unsigned long dummy) -{ - xpc_vars->heartbeat++; - - if (jiffies >= xpc_hb_check_timeout) { - wake_up_interruptible(&xpc_act_IRQ_wq); - } - - xpc_hb_timer.expires = jiffies + (xpc_hb_interval * HZ); - add_timer(&xpc_hb_timer); -} - - -/* - * This thread is responsible for nearly all of the partition - * activation/deactivation. - */ -static int -xpc_hb_checker(void *ignore) -{ - int last_IRQ_count = 0; - int new_IRQ_count; - int force_IRQ=0; - - - /* this thread was marked active by xpc_hb_init() */ - - daemonize(XPC_HB_CHECK_THREAD_NAME); - - set_cpus_allowed(current, cpumask_of_cpu(XPC_HB_CHECK_CPU)); - - xpc_hb_check_timeout = jiffies + (xpc_hb_check_interval * HZ); - - while (!(volatile int) xpc_exiting) { - - dev_dbg(xpc_part, "woke up with %d ticks rem; %d IRQs have " - "been received\n", - (int) (xpc_hb_check_timeout - jiffies), - atomic_read(&xpc_act_IRQ_rcvd) - last_IRQ_count); - - - /* checking of remote heartbeats is skewed by IRQ handling */ - if (jiffies >= xpc_hb_check_timeout) { - dev_dbg(xpc_part, "checking remote heartbeats\n"); - xpc_check_remote_hb(); - - /* - * We need to periodically recheck to ensure no - * IPI/AMO pairs have been missed. That check - * must always reset xpc_hb_check_timeout. - */ - force_IRQ = 1; - } - - - /* check for outstanding IRQs */ - new_IRQ_count = atomic_read(&xpc_act_IRQ_rcvd); - if (last_IRQ_count < new_IRQ_count || force_IRQ != 0) { - force_IRQ = 0; - - dev_dbg(xpc_part, "found an IRQ to process; will be " - "resetting xpc_hb_check_timeout\n"); - - last_IRQ_count += xpc_identify_act_IRQ_sender(); - if (last_IRQ_count < new_IRQ_count) { - /* retry once to help avoid missing AMO */ - (void) xpc_identify_act_IRQ_sender(); - } - last_IRQ_count = new_IRQ_count; - - xpc_hb_check_timeout = jiffies + - (xpc_hb_check_interval * HZ); - } - - /* wait for IRQ or timeout */ - (void) wait_event_interruptible(xpc_act_IRQ_wq, - (last_IRQ_count < atomic_read(&xpc_act_IRQ_rcvd) || - jiffies >= xpc_hb_check_timeout || - (volatile int) xpc_exiting)); - } - - dev_dbg(xpc_part, "heartbeat checker is exiting\n"); - - - /* mark this thread as having exited */ - complete(&xpc_hb_checker_exited); - return 0; -} - - -/* - * This thread will attempt to discover other partitions to activate - * based on info provided by SAL. This new thread is short lived and - * will exit once discovery is complete. - */ -static int -xpc_initiate_discovery(void *ignore) -{ - daemonize(XPC_DISCOVERY_THREAD_NAME); - - xpc_discovery(); - - dev_dbg(xpc_part, "discovery thread is exiting\n"); - - /* mark this thread as having exited */ - complete(&xpc_discovery_exited); - return 0; -} - - -/* - * Establish first contact with the remote partititon. This involves pulling - * the XPC per partition variables from the remote partition and waiting for - * the remote partition to pull ours. - */ -static enum xpc_retval -xpc_make_first_contact(struct xpc_partition *part) -{ - enum xpc_retval ret; - - - while ((ret = xpc_pull_remote_vars_part(part)) != xpcSuccess) { - if (ret != xpcRetry) { - XPC_DEACTIVATE_PARTITION(part, ret); - return ret; - } - - dev_dbg(xpc_chan, "waiting to make first contact with " - "partition %d\n", XPC_PARTID(part)); - - /* wait a 1/4 of a second or so */ - (void) msleep_interruptible(250); - - if (part->act_state == XPC_P_DEACTIVATING) { - return part->reason; - } - } - - return xpc_mark_partition_active(part); -} - - -/* - * The first kthread assigned to a newly activated partition is the one - * created by XPC HB with which it calls xpc_partition_up(). XPC hangs on to - * that kthread until the partition is brought down, at which time that kthread - * returns back to XPC HB. (The return of that kthread will signify to XPC HB - * that XPC has dismantled all communication infrastructure for the associated - * partition.) This kthread becomes the channel manager for that partition. - * - * Each active partition has a channel manager, who, besides connecting and - * disconnecting channels, will ensure that each of the partition's connected - * channels has the required number of assigned kthreads to get the work done. - */ -static void -xpc_channel_mgr(struct xpc_partition *part) -{ - while (part->act_state != XPC_P_DEACTIVATING || - atomic_read(&part->nchannels_active) > 0 || - !xpc_partition_disengaged(part)) { - - xpc_process_channel_activity(part); - - - /* - * Wait until we've been requested to activate kthreads or - * all of the channel's message queues have been torn down or - * a signal is pending. - * - * The channel_mgr_requests is set to 1 after being awakened, - * This is done to prevent the channel mgr from making one pass - * through the loop for each request, since he will - * be servicing all the requests in one pass. The reason it's - * set to 1 instead of 0 is so that other kthreads will know - * that the channel mgr is running and won't bother trying to - * wake him up. - */ - atomic_dec(&part->channel_mgr_requests); - (void) wait_event_interruptible(part->channel_mgr_wq, - (atomic_read(&part->channel_mgr_requests) > 0 || - (volatile u64) part->local_IPI_amo != 0 || - ((volatile u8) part->act_state == - XPC_P_DEACTIVATING && - atomic_read(&part->nchannels_active) == 0 && - xpc_partition_disengaged(part)))); - atomic_set(&part->channel_mgr_requests, 1); - - // >>> Does it need to wakeup periodically as well? In case we - // >>> miscalculated the #of kthreads to wakeup or create? - } -} - - -/* - * When XPC HB determines that a partition has come up, it will create a new - * kthread and that kthread will call this function to attempt to set up the - * basic infrastructure used for Cross Partition Communication with the newly - * upped partition. - * - * The kthread that was created by XPC HB and which setup the XPC - * infrastructure will remain assigned to the partition until the partition - * goes down. At which time the kthread will teardown the XPC infrastructure - * and then exit. - * - * XPC HB will put the remote partition's XPC per partition specific variables - * physical address into xpc_partitions[partid].remote_vars_part_pa prior to - * calling xpc_partition_up(). - */ -static void -xpc_partition_up(struct xpc_partition *part) -{ - DBUG_ON(part->channels != NULL); - - dev_dbg(xpc_chan, "activating partition %d\n", XPC_PARTID(part)); - - if (xpc_setup_infrastructure(part) != xpcSuccess) { - return; - } - - /* - * The kthread that XPC HB called us with will become the - * channel manager for this partition. It will not return - * back to XPC HB until the partition's XPC infrastructure - * has been dismantled. - */ - - (void) xpc_part_ref(part); /* this will always succeed */ - - if (xpc_make_first_contact(part) == xpcSuccess) { - xpc_channel_mgr(part); - } - - xpc_part_deref(part); - - xpc_teardown_infrastructure(part); -} - - -static int -xpc_activating(void *__partid) -{ - partid_t partid = (u64) __partid; - struct xpc_partition *part = &xpc_partitions[partid]; - unsigned long irq_flags; - struct sched_param param = { .sched_priority = MAX_RT_PRIO - 1 }; - int ret; - - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - - spin_lock_irqsave(&part->act_lock, irq_flags); - - if (part->act_state == XPC_P_DEACTIVATING) { - part->act_state = XPC_P_INACTIVE; - spin_unlock_irqrestore(&part->act_lock, irq_flags); - part->remote_rp_pa = 0; - return 0; - } - - /* indicate the thread is activating */ - DBUG_ON(part->act_state != XPC_P_ACTIVATION_REQ); - part->act_state = XPC_P_ACTIVATING; - - XPC_SET_REASON(part, 0, 0); - spin_unlock_irqrestore(&part->act_lock, irq_flags); - - dev_dbg(xpc_part, "bringing partition %d up\n", partid); - - daemonize("xpc%02d", partid); - - /* - * This thread needs to run at a realtime priority to prevent a - * significant performance degradation. - */ - ret = sched_setscheduler(current, SCHED_FIFO, ¶m); - if (ret != 0) { - dev_warn(xpc_part, "unable to set pid %d to a realtime " - "priority, ret=%d\n", current->pid, ret); - } - - /* allow this thread and its children to run on any CPU */ - set_cpus_allowed(current, CPU_MASK_ALL); - - /* - * Register the remote partition's AMOs with SAL so it can handle - * and cleanup errors within that address range should the remote - * partition go down. We don't unregister this range because it is - * difficult to tell when outstanding writes to the remote partition - * are finished and thus when it is safe to unregister. This should - * not result in wasted space in the SAL xp_addr_region table because - * we should get the same page for remote_amos_page_pa after module - * reloads and system reboots. - */ - if (sn_register_xp_addr_region(part->remote_amos_page_pa, - PAGE_SIZE, 1) < 0) { - dev_warn(xpc_part, "xpc_partition_up(%d) failed to register " - "xp_addr region\n", partid); - - spin_lock_irqsave(&part->act_lock, irq_flags); - part->act_state = XPC_P_INACTIVE; - XPC_SET_REASON(part, xpcPhysAddrRegFailed, __LINE__); - spin_unlock_irqrestore(&part->act_lock, irq_flags); - part->remote_rp_pa = 0; - return 0; - } - - xpc_allow_hb(partid, xpc_vars); - xpc_IPI_send_activated(part); - - - /* - * xpc_partition_up() holds this thread and marks this partition as - * XPC_P_ACTIVE by calling xpc_hb_mark_active(). - */ - (void) xpc_partition_up(part); - - xpc_disallow_hb(partid, xpc_vars); - xpc_mark_partition_inactive(part); - - if (part->reason == xpcReactivating) { - /* interrupting ourselves results in activating partition */ - xpc_IPI_send_reactivate(part); - } - - return 0; -} - - -void -xpc_activate_partition(struct xpc_partition *part) -{ - partid_t partid = XPC_PARTID(part); - unsigned long irq_flags; - pid_t pid; - - - spin_lock_irqsave(&part->act_lock, irq_flags); - - DBUG_ON(part->act_state != XPC_P_INACTIVE); - - part->act_state = XPC_P_ACTIVATION_REQ; - XPC_SET_REASON(part, xpcCloneKThread, __LINE__); - - spin_unlock_irqrestore(&part->act_lock, irq_flags); - - pid = kernel_thread(xpc_activating, (void *) ((u64) partid), 0); - - if (unlikely(pid <= 0)) { - spin_lock_irqsave(&part->act_lock, irq_flags); - part->act_state = XPC_P_INACTIVE; - XPC_SET_REASON(part, xpcCloneKThreadFailed, __LINE__); - spin_unlock_irqrestore(&part->act_lock, irq_flags); - } -} - - -/* - * Handle the receipt of a SGI_XPC_NOTIFY IRQ by seeing whether the specified - * partition actually sent it. Since SGI_XPC_NOTIFY IRQs may be shared by more - * than one partition, we use an AMO_t structure per partition to indicate - * whether a partition has sent an IPI or not. >>> If it has, then wake up the - * associated kthread to handle it. - * - * All SGI_XPC_NOTIFY IRQs received by XPC are the result of IPIs sent by XPC - * running on other partitions. - * - * Noteworthy Arguments: - * - * irq - Interrupt ReQuest number. NOT USED. - * - * dev_id - partid of IPI's potential sender. - * - * regs - processor's context before the processor entered - * interrupt code. NOT USED. - */ -irqreturn_t -xpc_notify_IRQ_handler(int irq, void *dev_id, struct pt_regs *regs) -{ - partid_t partid = (partid_t) (u64) dev_id; - struct xpc_partition *part = &xpc_partitions[partid]; - - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - - if (xpc_part_ref(part)) { - xpc_check_for_channel_activity(part); - - xpc_part_deref(part); - } - return IRQ_HANDLED; -} - - -/* - * Check to see if xpc_notify_IRQ_handler() dropped any IPIs on the floor - * because the write to their associated IPI amo completed after the IRQ/IPI - * was received. - */ -void -xpc_dropped_IPI_check(struct xpc_partition *part) -{ - if (xpc_part_ref(part)) { - xpc_check_for_channel_activity(part); - - part->dropped_IPI_timer.expires = jiffies + - XPC_P_DROPPED_IPI_WAIT; - add_timer(&part->dropped_IPI_timer); - xpc_part_deref(part); - } -} - - -void -xpc_activate_kthreads(struct xpc_channel *ch, int needed) -{ - int idle = atomic_read(&ch->kthreads_idle); - int assigned = atomic_read(&ch->kthreads_assigned); - int wakeup; - - - DBUG_ON(needed <= 0); - - if (idle > 0) { - wakeup = (needed > idle) ? idle : needed; - needed -= wakeup; - - dev_dbg(xpc_chan, "wakeup %d idle kthreads, partid=%d, " - "channel=%d\n", wakeup, ch->partid, ch->number); - - /* only wakeup the requested number of kthreads */ - wake_up_nr(&ch->idle_wq, wakeup); - } - - if (needed <= 0) { - return; - } - - if (needed + assigned > ch->kthreads_assigned_limit) { - needed = ch->kthreads_assigned_limit - assigned; - // >>>should never be less than 0 - if (needed <= 0) { - return; - } - } - - dev_dbg(xpc_chan, "create %d new kthreads, partid=%d, channel=%d\n", - needed, ch->partid, ch->number); - - xpc_create_kthreads(ch, needed); -} - - -/* - * This function is where XPC's kthreads wait for messages to deliver. - */ -static void -xpc_kthread_waitmsgs(struct xpc_partition *part, struct xpc_channel *ch) -{ - do { - /* deliver messages to their intended recipients */ - - while ((volatile s64) ch->w_local_GP.get < - (volatile s64) ch->w_remote_GP.put && - !((volatile u32) ch->flags & - XPC_C_DISCONNECTING)) { - xpc_deliver_msg(ch); - } - - if (atomic_inc_return(&ch->kthreads_idle) > - ch->kthreads_idle_limit) { - /* too many idle kthreads on this channel */ - atomic_dec(&ch->kthreads_idle); - break; - } - - dev_dbg(xpc_chan, "idle kthread calling " - "wait_event_interruptible_exclusive()\n"); - - (void) wait_event_interruptible_exclusive(ch->idle_wq, - ((volatile s64) ch->w_local_GP.get < - (volatile s64) ch->w_remote_GP.put || - ((volatile u32) ch->flags & - XPC_C_DISCONNECTING))); - - atomic_dec(&ch->kthreads_idle); - - } while (!((volatile u32) ch->flags & XPC_C_DISCONNECTING)); -} - - -static int -xpc_daemonize_kthread(void *args) -{ - partid_t partid = XPC_UNPACK_ARG1(args); - u16 ch_number = XPC_UNPACK_ARG2(args); - struct xpc_partition *part = &xpc_partitions[partid]; - struct xpc_channel *ch; - int n_needed; - unsigned long irq_flags; - - - daemonize("xpc%02dc%d", partid, ch_number); - - dev_dbg(xpc_chan, "kthread starting, partid=%d, channel=%d\n", - partid, ch_number); - - ch = &part->channels[ch_number]; - - if (!(ch->flags & XPC_C_DISCONNECTING)) { - - /* let registerer know that connection has been established */ - - spin_lock_irqsave(&ch->lock, irq_flags); - if (!(ch->flags & XPC_C_CONNECTEDCALLOUT)) { - ch->flags |= XPC_C_CONNECTEDCALLOUT; - spin_unlock_irqrestore(&ch->lock, irq_flags); - - xpc_connected_callout(ch); - - spin_lock_irqsave(&ch->lock, irq_flags); - ch->flags |= XPC_C_CONNECTEDCALLOUT_MADE; - spin_unlock_irqrestore(&ch->lock, irq_flags); - - /* - * It is possible that while the callout was being - * made that the remote partition sent some messages. - * If that is the case, we may need to activate - * additional kthreads to help deliver them. We only - * need one less than total #of messages to deliver. - */ - n_needed = ch->w_remote_GP.put - ch->w_local_GP.get - 1; - if (n_needed > 0 && - !(ch->flags & XPC_C_DISCONNECTING)) { - xpc_activate_kthreads(ch, n_needed); - } - } else { - spin_unlock_irqrestore(&ch->lock, irq_flags); - } - - xpc_kthread_waitmsgs(part, ch); - } - - if (atomic_dec_return(&ch->kthreads_assigned) == 0) { - spin_lock_irqsave(&ch->lock, irq_flags); - if ((ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) && - !(ch->flags & XPC_C_DISCONNECTINGCALLOUT)) { - ch->flags |= XPC_C_DISCONNECTINGCALLOUT; - spin_unlock_irqrestore(&ch->lock, irq_flags); - - xpc_disconnect_callout(ch, xpcDisconnecting); - - spin_lock_irqsave(&ch->lock, irq_flags); - ch->flags |= XPC_C_DISCONNECTINGCALLOUT_MADE; - } - spin_unlock_irqrestore(&ch->lock, irq_flags); - if (atomic_dec_return(&part->nchannels_engaged) == 0) { - xpc_mark_partition_disengaged(part); - xpc_IPI_send_disengage(part); - } - } - - - xpc_msgqueue_deref(ch); - - dev_dbg(xpc_chan, "kthread exiting, partid=%d, channel=%d\n", - partid, ch_number); - - xpc_part_deref(part); - return 0; -} - - -/* - * For each partition that XPC has established communications with, there is - * a minimum of one kernel thread assigned to perform any operation that - * may potentially sleep or block (basically the callouts to the asynchronous - * functions registered via xpc_connect()). - * - * Additional kthreads are created and destroyed by XPC as the workload - * demands. - * - * A kthread is assigned to one of the active channels that exists for a given - * partition. - */ -void -xpc_create_kthreads(struct xpc_channel *ch, int needed) -{ - unsigned long irq_flags; - pid_t pid; - u64 args = XPC_PACK_ARGS(ch->partid, ch->number); - struct xpc_partition *part = &xpc_partitions[ch->partid]; - - - while (needed-- > 0) { - - /* - * The following is done on behalf of the newly created - * kthread. That kthread is responsible for doing the - * counterpart to the following before it exits. - */ - (void) xpc_part_ref(part); - xpc_msgqueue_ref(ch); - if (atomic_inc_return(&ch->kthreads_assigned) == 1 && - atomic_inc_return(&part->nchannels_engaged) == 1) { - xpc_mark_partition_engaged(part); - } - - pid = kernel_thread(xpc_daemonize_kthread, (void *) args, 0); - if (pid < 0) { - /* the fork failed */ - if (atomic_dec_return(&ch->kthreads_assigned) == 0 && - atomic_dec_return(&part->nchannels_engaged) == 0) { - xpc_mark_partition_disengaged(part); - xpc_IPI_send_disengage(part); - } - xpc_msgqueue_deref(ch); - xpc_part_deref(part); - - if (atomic_read(&ch->kthreads_assigned) < - ch->kthreads_idle_limit) { - /* - * Flag this as an error only if we have an - * insufficient #of kthreads for the channel - * to function. - * - * No xpc_msgqueue_ref() is needed here since - * the channel mgr is doing this. - */ - spin_lock_irqsave(&ch->lock, irq_flags); - XPC_DISCONNECT_CHANNEL(ch, xpcLackOfResources, - &irq_flags); - spin_unlock_irqrestore(&ch->lock, irq_flags); - } - break; - } - - ch->kthreads_created++; // >>> temporary debug only!!! - } -} - - -void -xpc_disconnect_wait(int ch_number) -{ - unsigned long irq_flags; - partid_t partid; - struct xpc_partition *part; - struct xpc_channel *ch; - int wakeup_channel_mgr; - - - /* now wait for all callouts to the caller's function to cease */ - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - part = &xpc_partitions[partid]; - - if (!xpc_part_ref(part)) { - continue; - } - - ch = &part->channels[ch_number]; - - if (!(ch->flags & XPC_C_WDISCONNECT)) { - xpc_part_deref(part); - continue; - } - - wait_for_completion(&ch->wdisconnect_wait); - - spin_lock_irqsave(&ch->lock, irq_flags); - DBUG_ON(!(ch->flags & XPC_C_DISCONNECTED)); - wakeup_channel_mgr = 0; - - if (ch->delayed_IPI_flags) { - if (part->act_state != XPC_P_DEACTIVATING) { - spin_lock(&part->IPI_lock); - XPC_SET_IPI_FLAGS(part->local_IPI_amo, - ch->number, ch->delayed_IPI_flags); - spin_unlock(&part->IPI_lock); - wakeup_channel_mgr = 1; - } - ch->delayed_IPI_flags = 0; - } - - ch->flags &= ~XPC_C_WDISCONNECT; - spin_unlock_irqrestore(&ch->lock, irq_flags); - - if (wakeup_channel_mgr) { - xpc_wakeup_channel_mgr(part); - } - - xpc_part_deref(part); - } -} - - -static void -xpc_do_exit(enum xpc_retval reason) -{ - partid_t partid; - int active_part_count, printed_waiting_msg = 0; - struct xpc_partition *part; - unsigned long printmsg_time, disengage_request_timeout = 0; - - - /* a 'rmmod XPC' and a 'reboot' cannot both end up here together */ - DBUG_ON(xpc_exiting == 1); - - /* - * Let the heartbeat checker thread and the discovery thread - * (if one is running) know that they should exit. Also wake up - * the heartbeat checker thread in case it's sleeping. - */ - xpc_exiting = 1; - wake_up_interruptible(&xpc_act_IRQ_wq); - - /* ignore all incoming interrupts */ - free_irq(SGI_XPC_ACTIVATE, NULL); - - /* wait for the discovery thread to exit */ - wait_for_completion(&xpc_discovery_exited); - - /* wait for the heartbeat checker thread to exit */ - wait_for_completion(&xpc_hb_checker_exited); - - - /* sleep for a 1/3 of a second or so */ - (void) msleep_interruptible(300); - - - /* wait for all partitions to become inactive */ - - printmsg_time = jiffies + (XPC_DISENGAGE_PRINTMSG_INTERVAL * HZ); - xpc_disengage_request_timedout = 0; - - do { - active_part_count = 0; - - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - part = &xpc_partitions[partid]; - - if (xpc_partition_disengaged(part) && - part->act_state == XPC_P_INACTIVE) { - continue; - } - - active_part_count++; - - XPC_DEACTIVATE_PARTITION(part, reason); - - if (part->disengage_request_timeout > - disengage_request_timeout) { - disengage_request_timeout = - part->disengage_request_timeout; - } - } - - if (xpc_partition_engaged(-1UL)) { - if (time_after(jiffies, printmsg_time)) { - dev_info(xpc_part, "waiting for remote " - "partitions to disengage, timeout in " - "%ld seconds\n", - (disengage_request_timeout - jiffies) - / HZ); - printmsg_time = jiffies + - (XPC_DISENGAGE_PRINTMSG_INTERVAL * HZ); - printed_waiting_msg = 1; - } - - } else if (active_part_count > 0) { - if (printed_waiting_msg) { - dev_info(xpc_part, "waiting for local partition" - " to disengage\n"); - printed_waiting_msg = 0; - } - - } else { - if (!xpc_disengage_request_timedout) { - dev_info(xpc_part, "all partitions have " - "disengaged\n"); - } - break; - } - - /* sleep for a 1/3 of a second or so */ - (void) msleep_interruptible(300); - - } while (1); - - DBUG_ON(xpc_partition_engaged(-1UL)); - - - /* indicate to others that our reserved page is uninitialized */ - xpc_rsvd_page->vars_pa = 0; - - /* now it's time to eliminate our heartbeat */ - del_timer_sync(&xpc_hb_timer); - DBUG_ON(xpc_vars->heartbeating_to_mask != 0); - - if (reason == xpcUnloading) { - /* take ourselves off of the reboot_notifier_list */ - (void) unregister_reboot_notifier(&xpc_reboot_notifier); - - /* take ourselves off of the die_notifier list */ - (void) unregister_die_notifier(&xpc_die_notifier); - } - - /* close down protections for IPI operations */ - xpc_restrict_IPI_ops(); - - - /* clear the interface to XPC's functions */ - xpc_clear_interface(); - - if (xpc_sysctl) { - unregister_sysctl_table(xpc_sysctl); - } - - kfree(xpc_remote_copy_buffer_base); -} - - -/* - * This function is called when the system is being rebooted. - */ -static int -xpc_system_reboot(struct notifier_block *nb, unsigned long event, void *unused) -{ - enum xpc_retval reason; - - - switch (event) { - case SYS_RESTART: - reason = xpcSystemReboot; - break; - case SYS_HALT: - reason = xpcSystemHalt; - break; - case SYS_POWER_OFF: - reason = xpcSystemPoweroff; - break; - default: - reason = xpcSystemGoingDown; - } - - xpc_do_exit(reason); - return NOTIFY_DONE; -} - - -/* - * Notify other partitions to disengage from all references to our memory. - */ -static void -xpc_die_disengage(void) -{ - struct xpc_partition *part; - partid_t partid; - unsigned long engaged; - long time, printmsg_time, disengage_request_timeout; - - - /* keep xpc_hb_checker thread from doing anything (just in case) */ - xpc_exiting = 1; - - xpc_vars->heartbeating_to_mask = 0; /* indicate we're deactivated */ - - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - part = &xpc_partitions[partid]; - - if (!XPC_SUPPORTS_DISENGAGE_REQUEST(part-> - remote_vars_version)) { - - /* just in case it was left set by an earlier XPC */ - xpc_clear_partition_engaged(1UL << partid); - continue; - } - - if (xpc_partition_engaged(1UL << partid) || - part->act_state != XPC_P_INACTIVE) { - xpc_request_partition_disengage(part); - xpc_mark_partition_disengaged(part); - xpc_IPI_send_disengage(part); - } - } - - time = rtc_time(); - printmsg_time = time + - (XPC_DISENGAGE_PRINTMSG_INTERVAL * sn_rtc_cycles_per_second); - disengage_request_timeout = time + - (xpc_disengage_request_timelimit * sn_rtc_cycles_per_second); - - /* wait for all other partitions to disengage from us */ - - while (1) { - engaged = xpc_partition_engaged(-1UL); - if (!engaged) { - dev_info(xpc_part, "all partitions have disengaged\n"); - break; - } - - time = rtc_time(); - if (time >= disengage_request_timeout) { - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - if (engaged & (1UL << partid)) { - dev_info(xpc_part, "disengage from " - "remote partition %d timed " - "out\n", partid); - } - } - break; - } - - if (time >= printmsg_time) { - dev_info(xpc_part, "waiting for remote partitions to " - "disengage, timeout in %ld seconds\n", - (disengage_request_timeout - time) / - sn_rtc_cycles_per_second); - printmsg_time = time + - (XPC_DISENGAGE_PRINTMSG_INTERVAL * - sn_rtc_cycles_per_second); - } - } -} - - -/* - * This function is called when the system is being restarted or halted due - * to some sort of system failure. If this is the case we need to notify the - * other partitions to disengage from all references to our memory. - * This function can also be called when our heartbeater could be offlined - * for a time. In this case we need to notify other partitions to not worry - * about the lack of a heartbeat. - */ -static int -xpc_system_die(struct notifier_block *nb, unsigned long event, void *unused) -{ - switch (event) { - case DIE_MACHINE_RESTART: - case DIE_MACHINE_HALT: - xpc_die_disengage(); - break; - - case DIE_KDEBUG_ENTER: - /* Should lack of heartbeat be ignored by other partitions? */ - if (!xpc_kdebug_ignore) { - break; - } - /* fall through */ - case DIE_MCA_MONARCH_ENTER: - case DIE_INIT_MONARCH_ENTER: - xpc_vars->heartbeat++; - xpc_vars->heartbeat_offline = 1; - break; - - case DIE_KDEBUG_LEAVE: - /* Is lack of heartbeat being ignored by other partitions? */ - if (!xpc_kdebug_ignore) { - break; - } - /* fall through */ - case DIE_MCA_MONARCH_LEAVE: - case DIE_INIT_MONARCH_LEAVE: - xpc_vars->heartbeat++; - xpc_vars->heartbeat_offline = 0; - break; - } - - return NOTIFY_DONE; -} - - -int __init -xpc_init(void) -{ - int ret; - partid_t partid; - struct xpc_partition *part; - pid_t pid; - size_t buf_size; - - - if (!ia64_platform_is("sn2")) { - return -ENODEV; - } - - - buf_size = max(XPC_RP_VARS_SIZE, - XPC_RP_HEADER_SIZE + XP_NASID_MASK_BYTES); - xpc_remote_copy_buffer = xpc_kmalloc_cacheline_aligned(buf_size, - GFP_KERNEL, &xpc_remote_copy_buffer_base); - if (xpc_remote_copy_buffer == NULL) - return -ENOMEM; - - snprintf(xpc_part->bus_id, BUS_ID_SIZE, "part"); - snprintf(xpc_chan->bus_id, BUS_ID_SIZE, "chan"); - - xpc_sysctl = register_sysctl_table(xpc_sys_dir, 1); - - /* - * The first few fields of each entry of xpc_partitions[] need to - * be initialized now so that calls to xpc_connect() and - * xpc_disconnect() can be made prior to the activation of any remote - * partition. NOTE THAT NONE OF THE OTHER FIELDS BELONGING TO THESE - * ENTRIES ARE MEANINGFUL UNTIL AFTER AN ENTRY'S CORRESPONDING - * PARTITION HAS BEEN ACTIVATED. - */ - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - part = &xpc_partitions[partid]; - - DBUG_ON((u64) part != L1_CACHE_ALIGN((u64) part)); - - part->act_IRQ_rcvd = 0; - spin_lock_init(&part->act_lock); - part->act_state = XPC_P_INACTIVE; - XPC_SET_REASON(part, 0, 0); - - init_timer(&part->disengage_request_timer); - part->disengage_request_timer.function = - xpc_timeout_partition_disengage_request; - part->disengage_request_timer.data = (unsigned long) part; - - part->setup_state = XPC_P_UNSET; - init_waitqueue_head(&part->teardown_wq); - atomic_set(&part->references, 0); - } - - /* - * Open up protections for IPI operations (and AMO operations on - * Shub 1.1 systems). - */ - xpc_allow_IPI_ops(); - - /* - * Interrupts being processed will increment this atomic variable and - * awaken the heartbeat thread which will process the interrupts. - */ - atomic_set(&xpc_act_IRQ_rcvd, 0); - - /* - * This is safe to do before the xpc_hb_checker thread has started - * because the handler releases a wait queue. If an interrupt is - * received before the thread is waiting, it will not go to sleep, - * but rather immediately process the interrupt. - */ - ret = request_irq(SGI_XPC_ACTIVATE, xpc_act_IRQ_handler, 0, - "xpc hb", NULL); - if (ret != 0) { - dev_err(xpc_part, "can't register ACTIVATE IRQ handler, " - "errno=%d\n", -ret); - - xpc_restrict_IPI_ops(); - - if (xpc_sysctl) { - unregister_sysctl_table(xpc_sysctl); - } - - kfree(xpc_remote_copy_buffer_base); - return -EBUSY; - } - - /* - * Fill the partition reserved page with the information needed by - * other partitions to discover we are alive and establish initial - * communications. - */ - xpc_rsvd_page = xpc_rsvd_page_init(); - if (xpc_rsvd_page == NULL) { - dev_err(xpc_part, "could not setup our reserved page\n"); - - free_irq(SGI_XPC_ACTIVATE, NULL); - xpc_restrict_IPI_ops(); - - if (xpc_sysctl) { - unregister_sysctl_table(xpc_sysctl); - } - - kfree(xpc_remote_copy_buffer_base); - return -EBUSY; - } - - - /* add ourselves to the reboot_notifier_list */ - ret = register_reboot_notifier(&xpc_reboot_notifier); - if (ret != 0) { - dev_warn(xpc_part, "can't register reboot notifier\n"); - } - - /* add ourselves to the die_notifier list (i.e., ia64die_chain) */ - ret = register_die_notifier(&xpc_die_notifier); - if (ret != 0) { - dev_warn(xpc_part, "can't register die notifier\n"); - } - - - /* - * Set the beating to other partitions into motion. This is - * the last requirement for other partitions' discovery to - * initiate communications with us. - */ - init_timer(&xpc_hb_timer); - xpc_hb_timer.function = xpc_hb_beater; - xpc_hb_beater(0); - - - /* - * The real work-horse behind xpc. This processes incoming - * interrupts and monitors remote heartbeats. - */ - pid = kernel_thread(xpc_hb_checker, NULL, 0); - if (pid < 0) { - dev_err(xpc_part, "failed while forking hb check thread\n"); - - /* indicate to others that our reserved page is uninitialized */ - xpc_rsvd_page->vars_pa = 0; - - /* take ourselves off of the reboot_notifier_list */ - (void) unregister_reboot_notifier(&xpc_reboot_notifier); - - /* take ourselves off of the die_notifier list */ - (void) unregister_die_notifier(&xpc_die_notifier); - - del_timer_sync(&xpc_hb_timer); - free_irq(SGI_XPC_ACTIVATE, NULL); - xpc_restrict_IPI_ops(); - - if (xpc_sysctl) { - unregister_sysctl_table(xpc_sysctl); - } - - kfree(xpc_remote_copy_buffer_base); - return -EBUSY; - } - - - /* - * Startup a thread that will attempt to discover other partitions to - * activate based on info provided by SAL. This new thread is short - * lived and will exit once discovery is complete. - */ - pid = kernel_thread(xpc_initiate_discovery, NULL, 0); - if (pid < 0) { - dev_err(xpc_part, "failed while forking discovery thread\n"); - - /* mark this new thread as a non-starter */ - complete(&xpc_discovery_exited); - - xpc_do_exit(xpcUnloading); - return -EBUSY; - } - - - /* set the interface to point at XPC's functions */ - xpc_set_interface(xpc_initiate_connect, xpc_initiate_disconnect, - xpc_initiate_allocate, xpc_initiate_send, - xpc_initiate_send_notify, xpc_initiate_received, - xpc_initiate_partid_to_nasids); - - return 0; -} -module_init(xpc_init); - - -void __exit -xpc_exit(void) -{ - xpc_do_exit(xpcUnloading); -} -module_exit(xpc_exit); - - -MODULE_AUTHOR("Silicon Graphics, Inc."); -MODULE_DESCRIPTION("Cross Partition Communication (XPC) support"); -MODULE_LICENSE("GPL"); - -module_param(xpc_hb_interval, int, 0); -MODULE_PARM_DESC(xpc_hb_interval, "Number of seconds between " - "heartbeat increments."); - -module_param(xpc_hb_check_interval, int, 0); -MODULE_PARM_DESC(xpc_hb_check_interval, "Number of seconds between " - "heartbeat checks."); - -module_param(xpc_disengage_request_timelimit, int, 0); -MODULE_PARM_DESC(xpc_disengage_request_timelimit, "Number of seconds to wait " - "for disengage request to complete."); - -module_param(xpc_kdebug_ignore, int, 0); -MODULE_PARM_DESC(xpc_kdebug_ignore, "Should lack of heartbeat be ignored by " - "other partitions when dropping into kdebug."); - diff --git a/arch/ia64/sn/kernel/xpc_partition.c b/arch/ia64/sn/kernel/xpc_partition.c deleted file mode 100644 index 57c723f5cba..00000000000 --- a/arch/ia64/sn/kernel/xpc_partition.c +++ /dev/null @@ -1,1239 +0,0 @@ -/* - * This file is subject to the terms and conditions of the GNU General Public - * License. See the file "COPYING" in the main directory of this archive - * for more details. - * - * Copyright (c) 2004-2006 Silicon Graphics, Inc. All Rights Reserved. - */ - - -/* - * Cross Partition Communication (XPC) partition support. - * - * This is the part of XPC that detects the presence/absence of - * other partitions. It provides a heartbeat and monitors the - * heartbeats of other partitions. - * - */ - - -#include <linux/kernel.h> -#include <linux/sysctl.h> -#include <linux/cache.h> -#include <linux/mmzone.h> -#include <linux/nodemask.h> -#include <asm/uncached.h> -#include <asm/sn/bte.h> -#include <asm/sn/intr.h> -#include <asm/sn/sn_sal.h> -#include <asm/sn/nodepda.h> -#include <asm/sn/addrs.h> -#include <asm/sn/xpc.h> - - -/* XPC is exiting flag */ -int xpc_exiting; - - -/* SH_IPI_ACCESS shub register value on startup */ -static u64 xpc_sh1_IPI_access; -static u64 xpc_sh2_IPI_access0; -static u64 xpc_sh2_IPI_access1; -static u64 xpc_sh2_IPI_access2; -static u64 xpc_sh2_IPI_access3; - - -/* original protection values for each node */ -u64 xpc_prot_vec[MAX_NUMNODES]; - - -/* this partition's reserved page pointers */ -struct xpc_rsvd_page *xpc_rsvd_page; -static u64 *xpc_part_nasids; -static u64 *xpc_mach_nasids; -struct xpc_vars *xpc_vars; -struct xpc_vars_part *xpc_vars_part; - -static int xp_nasid_mask_bytes; /* actual size in bytes of nasid mask */ -static int xp_nasid_mask_words; /* actual size in words of nasid mask */ - - -/* - * For performance reasons, each entry of xpc_partitions[] is cacheline - * aligned. And xpc_partitions[] is padded with an additional entry at the - * end so that the last legitimate entry doesn't share its cacheline with - * another variable. - */ -struct xpc_partition xpc_partitions[XP_MAX_PARTITIONS + 1]; - - -/* - * Generic buffer used to store a local copy of portions of a remote - * partition's reserved page (either its header and part_nasids mask, - * or its vars). - */ -char *xpc_remote_copy_buffer; -void *xpc_remote_copy_buffer_base; - - -/* - * Guarantee that the kmalloc'd memory is cacheline aligned. - */ -void * -xpc_kmalloc_cacheline_aligned(size_t size, gfp_t flags, void **base) -{ - /* see if kmalloc will give us cachline aligned memory by default */ - *base = kmalloc(size, flags); - if (*base == NULL) { - return NULL; - } - if ((u64) *base == L1_CACHE_ALIGN((u64) *base)) { - return *base; - } - kfree(*base); - - /* nope, we'll have to do it ourselves */ - *base = kmalloc(size + L1_CACHE_BYTES, flags); - if (*base == NULL) { - return NULL; - } - return (void *) L1_CACHE_ALIGN((u64) *base); -} - - -/* - * Given a nasid, get the physical address of the partition's reserved page - * for that nasid. This function returns 0 on any error. - */ -static u64 -xpc_get_rsvd_page_pa(int nasid) -{ - bte_result_t bte_res; - s64 status; - u64 cookie = 0; - u64 rp_pa = nasid; /* seed with nasid */ - u64 len = 0; - u64 buf = buf; - u64 buf_len = 0; - void *buf_base = NULL; - - - while (1) { - - status = sn_partition_reserved_page_pa(buf, &cookie, &rp_pa, - &len); - - dev_dbg(xpc_part, "SAL returned with status=%li, cookie=" - "0x%016lx, address=0x%016lx, len=0x%016lx\n", - status, cookie, rp_pa, len); - - if (status != SALRET_MORE_PASSES) { - break; - } - - if (L1_CACHE_ALIGN(len) > buf_len) { - kfree(buf_base); - buf_len = L1_CACHE_ALIGN(len); - buf = (u64) xpc_kmalloc_cacheline_aligned(buf_len, - GFP_KERNEL, &buf_base); - if (buf_base == NULL) { - dev_err(xpc_part, "unable to kmalloc " - "len=0x%016lx\n", buf_len); - status = SALRET_ERROR; - break; - } - } - - bte_res = xp_bte_copy(rp_pa, buf, buf_len, - (BTE_NOTIFY | BTE_WACQUIRE), NULL); - if (bte_res != BTE_SUCCESS) { - dev_dbg(xpc_part, "xp_bte_copy failed %i\n", bte_res); - status = SALRET_ERROR; - break; - } - } - - kfree(buf_base); - - if (status != SALRET_OK) { - rp_pa = 0; - } - dev_dbg(xpc_part, "reserved page at phys address 0x%016lx\n", rp_pa); - return rp_pa; -} - - -/* - * Fill the partition reserved page with the information needed by - * other partitions to discover we are alive and establish initial - * communications. - */ -struct xpc_rsvd_page * -xpc_rsvd_page_init(void) -{ - struct xpc_rsvd_page *rp; - AMO_t *amos_page; - u64 rp_pa, nasid_array = 0; - int i, ret; - - - /* get the local reserved page's address */ - - preempt_disable(); - rp_pa = xpc_get_rsvd_page_pa(cpuid_to_nasid(smp_processor_id())); - preempt_enable(); - if (rp_pa == 0) { - dev_err(xpc_part, "SAL failed to locate the reserved page\n"); - return NULL; - } - rp = (struct xpc_rsvd_page *) __va(rp_pa); - - if (rp->partid != sn_partition_id) { - dev_err(xpc_part, "the reserved page's partid of %d should be " - "%d\n", rp->partid, sn_partition_id); - return NULL; - } - - rp->version = XPC_RP_VERSION; - - /* establish the actual sizes of the nasid masks */ - if (rp->SAL_version == 1) { - /* SAL_version 1 didn't set the nasids_size field */ - rp->nasids_size = 128; - } - xp_nasid_mask_bytes = rp->nasids_size; - xp_nasid_mask_words = xp_nasid_mask_bytes / 8; - - /* setup the pointers to the various items in the reserved page */ - xpc_part_nasids = XPC_RP_PART_NASIDS(rp); - xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp); - xpc_vars = XPC_RP_VARS(rp); - xpc_vars_part = XPC_RP_VARS_PART(rp); - - /* - * Before clearing xpc_vars, see if a page of AMOs had been previously - * allocated. If not we'll need to allocate one and set permissions - * so that cross-partition AMOs are allowed. - * - * The allocated AMO page needs MCA reporting to remain disabled after - * XPC has unloaded. To make this work, we keep a copy of the pointer - * to this page (i.e., amos_page) in the struct xpc_vars structure, - * which is pointed to by the reserved page, and re-use that saved copy - * on subsequent loads of XPC. This AMO page is never freed, and its - * memory protections are never restricted. - */ - if ((amos_page = xpc_vars->amos_page) == NULL) { - amos_page = (AMO_t *) TO_AMO(uncached_alloc_page(0)); - if (amos_page == NULL) { - dev_err(xpc_part, "can't allocate page of AMOs\n"); - return NULL; - } - - /* - * Open up AMO-R/W to cpu. This is done for Shub 1.1 systems - * when xpc_allow_IPI_ops() is called via xpc_hb_init(). - */ - if (!enable_shub_wars_1_1()) { - ret = sn_change_memprotect(ia64_tpa((u64) amos_page), - PAGE_SIZE, SN_MEMPROT_ACCESS_CLASS_1, - &nasid_array); - if (ret != 0) { - dev_err(xpc_part, "can't change memory " - "protections\n"); - uncached_free_page(__IA64_UNCACHED_OFFSET | - TO_PHYS((u64) amos_page)); - return NULL; - } - } - } else if (!IS_AMO_ADDRESS((u64) amos_page)) { - /* - * EFI's XPBOOT can also set amos_page in the reserved page, - * but it happens to leave it as an uncached physical address - * and we need it to be an uncached virtual, so we'll have to - * convert it. - */ - if (!IS_AMO_PHYS_ADDRESS((u64) amos_page)) { - dev_err(xpc_part, "previously used amos_page address " - "is bad = 0x%p\n", (void *) amos_page); - return NULL; - } - amos_page = (AMO_t *) TO_AMO((u64) amos_page); - } - - /* clear xpc_vars */ - memset(xpc_vars, 0, sizeof(struct xpc_vars)); - - xpc_vars->version = XPC_V_VERSION; - xpc_vars->act_nasid = cpuid_to_nasid(0); - xpc_vars->act_phys_cpuid = cpu_physical_id(0); - xpc_vars->vars_part_pa = __pa(xpc_vars_part); - xpc_vars->amos_page_pa = ia64_tpa((u64) amos_page); - xpc_vars->amos_page = amos_page; /* save for next load of XPC */ - - - /* clear xpc_vars_part */ - memset((u64 *) xpc_vars_part, 0, sizeof(struct xpc_vars_part) * - XP_MAX_PARTITIONS); - - /* initialize the activate IRQ related AMO variables */ - for (i = 0; i < xp_nasid_mask_words; i++) { - (void) xpc_IPI_init(XPC_ACTIVATE_IRQ_AMOS + i); - } - - /* initialize the engaged remote partitions related AMO variables */ - (void) xpc_IPI_init(XPC_ENGAGED_PARTITIONS_AMO); - (void) xpc_IPI_init(XPC_DISENGAGE_REQUEST_AMO); - - /* timestamp of when reserved page was setup by XPC */ - rp->stamp = CURRENT_TIME; - - /* - * This signifies to the remote partition that our reserved - * page is initialized. - */ - rp->vars_pa = __pa(xpc_vars); - - return rp; -} - - -/* - * Change protections to allow IPI operations (and AMO operations on - * Shub 1.1 systems). - */ -void -xpc_allow_IPI_ops(void) -{ - int node; - int nasid; - - - // >>> Change SH_IPI_ACCESS code to use SAL call once it is available. - - if (is_shub2()) { - xpc_sh2_IPI_access0 = - (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS0)); - xpc_sh2_IPI_access1 = - (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS1)); - xpc_sh2_IPI_access2 = - (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS2)); - xpc_sh2_IPI_access3 = - (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS3)); - - for_each_online_node(node) { - nasid = cnodeid_to_nasid(node); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0), - -1UL); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1), - -1UL); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2), - -1UL); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3), - -1UL); - } - - } else { - xpc_sh1_IPI_access = - (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH1_IPI_ACCESS)); - - for_each_online_node(node) { - nasid = cnodeid_to_nasid(node); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS), - -1UL); - - /* - * Since the BIST collides with memory operations on - * SHUB 1.1 sn_change_memprotect() cannot be used. - */ - if (enable_shub_wars_1_1()) { - /* open up everything */ - xpc_prot_vec[node] = (u64) HUB_L((u64 *) - GLOBAL_MMR_ADDR(nasid, - SH1_MD_DQLP_MMR_DIR_PRIVEC0)); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, - SH1_MD_DQLP_MMR_DIR_PRIVEC0), - -1UL); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, - SH1_MD_DQRP_MMR_DIR_PRIVEC0), - -1UL); - } - } - } -} - - -/* - * Restrict protections to disallow IPI operations (and AMO operations on - * Shub 1.1 systems). - */ -void -xpc_restrict_IPI_ops(void) -{ - int node; - int nasid; - - - // >>> Change SH_IPI_ACCESS code to use SAL call once it is available. - - if (is_shub2()) { - - for_each_online_node(node) { - nasid = cnodeid_to_nasid(node); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0), - xpc_sh2_IPI_access0); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1), - xpc_sh2_IPI_access1); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2), - xpc_sh2_IPI_access2); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3), - xpc_sh2_IPI_access3); - } - - } else { - - for_each_online_node(node) { - nasid = cnodeid_to_nasid(node); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS), - xpc_sh1_IPI_access); - - if (enable_shub_wars_1_1()) { - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, - SH1_MD_DQLP_MMR_DIR_PRIVEC0), - xpc_prot_vec[node]); - HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, - SH1_MD_DQRP_MMR_DIR_PRIVEC0), - xpc_prot_vec[node]); - } - } - } -} - - -/* - * At periodic intervals, scan through all active partitions and ensure - * their heartbeat is still active. If not, the partition is deactivated. - */ -void -xpc_check_remote_hb(void) -{ - struct xpc_vars *remote_vars; - struct xpc_partition *part; - partid_t partid; - bte_result_t bres; - - - remote_vars = (struct xpc_vars *) xpc_remote_copy_buffer; - - for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) { - - if (xpc_exiting) { - break; - } - - if (partid == sn_partition_id) { - continue; - } - - part = &xpc_partitions[partid]; - - if (part->act_state == XPC_P_INACTIVE || - part->act_state == XPC_P_DEACTIVATING) { - continue; - } - - /* pull the remote_hb cache line */ - bres = xp_bte_copy(part->remote_vars_pa, - (u64) remote_vars, - XPC_RP_VARS_SIZE, - (BTE_NOTIFY | BTE_WACQUIRE), NULL); - if (bres != BTE_SUCCESS) { - XPC_DEACTIVATE_PARTITION(part, - xpc_map_bte_errors(bres)); - continue; - } - - dev_dbg(xpc_part, "partid = %d, heartbeat = %ld, last_heartbeat" - " = %ld, heartbeat_offline = %ld, HB_mask = 0x%lx\n", - partid, remote_vars->heartbeat, part->last_heartbeat, - remote_vars->heartbeat_offline, - remote_vars->heartbeating_to_mask); - - if (((remote_vars->heartbeat == part->last_heartbeat) && - (remote_vars->heartbeat_offline == 0)) || - !xpc_hb_allowed(sn_partition_id, remote_vars)) { - - XPC_DEACTIVATE_PARTITION(part, xpcNoHeartbeat); - continue; - } - - part->last_heartbeat = remote_vars->heartbeat; - } -} - - -/* - * Get a copy of a portion of the remote partition's rsvd page. - * - * remote_rp points to a buffer that is cacheline aligned for BTE copies and - * is large enough to contain a copy of their reserved page header and - * part_nasids mask. - */ -static enum xpc_retval -xpc_get_remote_rp(int nasid, u64 *discovered_nasids, - struct xpc_rsvd_page *remote_rp, u64 *remote_rp_pa) -{ - int bres, i; - - - /* get the reserved page's physical address */ - - *remote_rp_pa = xpc_get_rsvd_page_pa(nasid); - if (*remote_rp_pa == 0) { - return xpcNoRsvdPageAddr; - } - - - /* pull over the reserved page header and part_nasids mask */ - bres = xp_bte_copy(*remote_rp_pa, (u64) remote_rp, - XPC_RP_HEADER_SIZE + xp_nasid_mask_bytes, - (BTE_NOTIFY | BTE_WACQUIRE), NULL); - if (bres != BTE_SUCCESS) { - return xpc_map_bte_errors(bres); - } - - - if (discovered_nasids != NULL) { - u64 *remote_part_nasids = XPC_RP_PART_NASIDS(remote_rp); - - - for (i = 0; i < xp_nasid_mask_words; i++) { - discovered_nasids[i] |= remote_part_nasids[i]; - } - } - - - /* check that the partid is for another partition */ - - if (remote_rp->partid < 1 || - remote_rp->partid > (XP_MAX_PARTITIONS - 1)) { - return xpcInvalidPartid; - } - - if (remote_rp->partid == sn_partition_id) { - return xpcLocalPartid; - } - - - if (XPC_VERSION_MAJOR(remote_rp->version) != - XPC_VERSION_MAJOR(XPC_RP_VERSION)) { - return xpcBadVersion; - } - - return xpcSuccess; -} - - -/* - * Get a copy of the remote partition's XPC variables from the reserved page. - * - * remote_vars points to a buffer that is cacheline aligned for BTE copies and - * assumed to be of size XPC_RP_VARS_SIZE. - */ -static enum xpc_retval -xpc_get_remote_vars(u64 remote_vars_pa, struct xpc_vars *remote_vars) -{ - int bres; - - - if (remote_vars_pa == 0) { - return xpcVarsNotSet; - } - - /* pull over the cross partition variables */ - bres = xp_bte_copy(remote_vars_pa, (u64) remote_vars, XPC_RP_VARS_SIZE, - (BTE_NOTIFY | BTE_WACQUIRE), NULL); - if (bres != BTE_SUCCESS) { - return xpc_map_bte_errors(bres); - } - - if (XPC_VERSION_MAJOR(remote_vars->version) != - XPC_VERSION_MAJOR(XPC_V_VERSION)) { - return xpcBadVersion; - } - - return xpcSuccess; -} - - -/* - * Update the remote partition's info. - */ -static void -xpc_update_partition_info(struct xpc_partition *part, u8 remote_rp_version, - struct timespec *remote_rp_stamp, u64 remote_rp_pa, - u64 remote_vars_pa, struct xpc_vars *remote_vars) -{ - part->remote_rp_version = remote_rp_version; - dev_dbg(xpc_part, " remote_rp_version = 0x%016lx\n", - part->remote_rp_version); - - part->remote_rp_stamp = *remote_rp_stamp; - dev_dbg(xpc_part, " remote_rp_stamp (tv_sec = 0x%lx tv_nsec = 0x%lx\n", - part->remote_rp_stamp.tv_sec, part->remote_rp_stamp.tv_nsec); - - part->remote_rp_pa = remote_rp_pa; - dev_dbg(xpc_part, " remote_rp_pa = 0x%016lx\n", part->remote_rp_pa); - - part->remote_vars_pa = remote_vars_pa; - dev_dbg(xpc_part, " remote_vars_pa = 0x%016lx\n", - part->remote_vars_pa); - - part->last_heartbeat = remote_vars->heartbeat; - dev_dbg(xpc_part, " last_heartbeat = 0x%016lx\n", - part->last_heartbeat); - - part->remote_vars_part_pa = remote_vars->vars_part_pa; - dev_dbg(xpc_part, " remote_vars_part_pa = 0x%016lx\n", - part->remote_vars_part_pa); - - part->remote_act_nasid = remote_vars->act_nasid; - dev_dbg(xpc_part, " remote_act_nasid = 0x%x\n", - part->remote_act_nasid); - - part->remote_act_phys_cpuid = remote_vars->act_phys_cpuid; - dev_dbg(xpc_part, " remote_act_phys_cpuid = 0x%x\n", - part->remote_act_phys_cpuid); - - part->remote_amos_page_pa = remote_vars->amos_page_pa; - dev_dbg(xpc_part, " remote_amos_page_pa = 0x%lx\n", - part->remote_amos_page_pa); - - part->remote_vars_version = remote_vars->version; - dev_dbg(xpc_part, " remote_vars_version = 0x%x\n", - part->remote_vars_version); -} - - -/* - * Prior code has determined the nasid which generated an IPI. Inspect - * that nasid to determine if its partition needs to be activated or - * deactivated. - * - * A partition is consider "awaiting activation" if our partition - * flags indicate it is not active and it has a heartbeat. A - * partition is considered "awaiting deactivation" if our partition - * flags indicate it is active but it has no heartbeat or it is not - * sending its heartbeat to us. - * - * To determine the heartbeat, the remote nasid must have a properly - * initialized reserved page. - */ -static void -xpc_identify_act_IRQ_req(int nasid) -{ - struct xpc_rsvd_page *remote_rp; - struct xpc_vars *remote_vars; - u64 remote_rp_pa; - u64 remote_vars_pa; - int remote_rp_version; - int reactivate = 0; - int stamp_diff; - struct timespec remote_rp_stamp = { 0, 0 }; - partid_t partid; - struct xpc_partition *part; - enum xpc_retval ret; - - - /* pull over the reserved page structure */ - - remote_rp = (struct xpc_rsvd_page *) xpc_remote_copy_buffer; - - ret = xpc_get_remote_rp(nasid, NULL, remote_rp, &remote_rp_pa); - if (ret != xpcSuccess) { - dev_warn(xpc_part, "unable to get reserved page from nasid %d, " - "which sent interrupt, reason=%d\n", nasid, ret); - return; - } - - remote_vars_pa = remote_rp->vars_pa; - remote_rp_version = remote_rp->version; - if (XPC_SUPPORTS_RP_STAMP(remote_rp_version)) { - remote_rp_stamp = remote_rp->stamp; - } - partid = remote_rp->partid; - part = &xpc_partitions[partid]; - - - /* pull over the cross partition variables */ - - remote_vars = (struct xpc_vars *) xpc_remote_copy_buffer; - - ret = xpc_get_remote_vars(remote_vars_pa, remote_vars); - if (ret != xpcSuccess) { - - dev_warn(xpc_part, "unable to get XPC variables from nasid %d, " - "which sent interrupt, reason=%d\n", nasid, ret); - - XPC_DEACTIVATE_PARTITION(part, ret); - return; - } - - - part->act_IRQ_rcvd++; - - dev_dbg(xpc_part, "partid for nasid %d is %d; IRQs = %d; HB = " - "%ld:0x%lx\n", (int) nasid, (int) partid, part->act_IRQ_rcvd, - remote_vars->heartbeat, remote_vars->heartbeating_to_mask); - - if (xpc_partition_disengaged(part) && - part->act_state == XPC_P_INACTIVE) { - - xpc_update_partition_info(part, remote_rp_version, - &remote_rp_stamp, remote_rp_pa, - remote_vars_pa, remote_vars); - - if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) { - if (xpc_partition_disengage_requested(1UL << partid)) { - /* - * Other side is waiting on us to disengage, - * even though we already have. - */ - return; - } - } else { - /* other side doesn't support disengage requests */ - xpc_clear_partition_disengage_request(1UL << partid); - } - - xpc_activate_partition(part); - return; - } - - DBUG_ON(part->remote_rp_version == 0); - DBUG_ON(part->remote_vars_version == 0); - - if (!XPC_SUPPORTS_RP_STAMP(part->remote_rp_version)) { - DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(part-> - remote_vars_version)); - - if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) { - DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars-> - version)); - /* see if the other side rebooted */ - if (part->remote_amos_page_pa == - remote_vars->amos_page_pa && - xpc_hb_allowed(sn_partition_id, - remote_vars)) { - /* doesn't look that way, so ignore the IPI */ - return; - } - } - - /* - * Other side rebooted and previous XPC didn't support the - * disengage request, so we don't need to do anything special. - */ - - xpc_update_partition_info(part, remote_rp_version, - &remote_rp_stamp, remote_rp_pa, - remote_vars_pa, remote_vars); - part->reactivate_nasid = nasid; - XPC_DEACTIVATE_PARTITION(part, xpcReactivating); - return; - } - - DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)); - - if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) { - DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version)); - - /* - * Other side rebooted and previous XPC did support the - * disengage request, but the new one doesn't. - */ - - xpc_clear_partition_engaged(1UL << partid); - xpc_clear_partition_disengage_request(1UL << partid); - - xpc_update_partition_info(part, remote_rp_version, - &remote_rp_stamp, remote_rp_pa, - remote_vars_pa, remote_vars); - reactivate = 1; - - } else { - DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version)); - - stamp_diff = xpc_compare_stamps(&part->remote_rp_stamp, - &remote_rp_stamp); - if (stamp_diff != 0) { - DBUG_ON(stamp_diff >= 0); - - /* - * Other side rebooted and the previous XPC did support - * the disengage request, as does the new one. - */ - - DBUG_ON(xpc_partition_engaged(1UL << partid)); - DBUG_ON(xpc_partition_disengage_requested(1UL << - partid)); - - xpc_update_partition_info(part, remote_rp_version, - &remote_rp_stamp, remote_rp_pa, - remote_vars_pa, remote_vars); - reactivate = 1; - } - } - - if (part->disengage_request_timeout > 0 && - !xpc_partition_disengaged(part)) { - /* still waiting on other side to disengage from us */ - return; - } - - if (reactivate) { - part->reactivate_nasid = nasid; - XPC_DEACTIVATE_PARTITION(part, xpcReactivating); - - } else if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version) && - xpc_partition_disengage_requested(1UL << partid)) { - XPC_DEACTIVATE_PARTITION(part, xpcOtherGoingDown); - } -} - - -/* - * Loop through the activation AMO variables and process any bits - * which are set. Each bit indicates a nasid sending a partition - * activation or deactivation request. - * - * Return #of IRQs detected. - */ -int -xpc_identify_act_IRQ_sender(void) -{ - int word, bit; - u64 nasid_mask; - u64 nasid; /* remote nasid */ - int n_IRQs_detected = 0; - AMO_t *act_amos; - - - act_amos = xpc_vars->amos_page + XPC_ACTIVATE_IRQ_AMOS; - - - /* scan through act AMO variable looking for non-zero entries */ - for (word = 0; word < xp_nasid_mask_words; word++) { - - if (xpc_exiting) { - break; - } - - nasid_mask = xpc_IPI_receive(&act_amos[word]); - if (nasid_mask == 0) { - /* no IRQs from nasids in this variable */ - continue; - } - - dev_dbg(xpc_part, "AMO[%d] gave back 0x%lx\n", word, - nasid_mask); - - - /* - * If this nasid has been added to the machine since - * our partition was reset, this will retain the - * remote nasid in our reserved pages machine mask. - * This is used in the event of module reload. - */ - xpc_mach_nasids[word] |= nasid_mask; - - - /* locate the nasid(s) which sent interrupts */ - - for (bit = 0; bit < (8 * sizeof(u64)); bit++) { - if (nasid_mask & (1UL << bit)) { - n_IRQs_detected++; - nasid = XPC_NASID_FROM_W_B(word, bit); - dev_dbg(xpc_part, "interrupt from nasid %ld\n", - nasid); - xpc_identify_act_IRQ_req(nasid); - } - } - } - return n_IRQs_detected; -} - - -/* - * See if the other side has responded to a partition disengage request - * from us. - */ -int -xpc_partition_disengaged(struct xpc_partition *part) -{ - partid_t partid = XPC_PARTID(part); - int disengaged; - - - disengaged = (xpc_partition_engaged(1UL << partid) == 0); - if (part->disengage_request_timeout) { - if (!disengaged) { - if (jiffies < part->disengage_request_timeout) { - /* timelimit hasn't been reached yet */ - return 0; - } - - /* - * Other side hasn't responded to our disengage - * request in a timely fashion, so assume it's dead. - */ - - dev_info(xpc_part, "disengage from remote partition %d " - "timed out\n", partid); - xpc_disengage_request_timedout = 1; - xpc_clear_partition_engaged(1UL << partid); - disengaged = 1; - } - part->disengage_request_timeout = 0; - - /* cancel the timer function, provided it's not us */ - if (!in_interrupt()) { - del_singleshot_timer_sync(&part-> - disengage_request_timer); - } - - DBUG_ON(part->act_state != XPC_P_DEACTIVATING && - part->act_state != XPC_P_INACTIVE); - if (part->act_state != XPC_P_INACTIVE) { - xpc_wakeup_channel_mgr(part); - } - - if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) { - xpc_cancel_partition_disengage_request(part); - } - } - return disengaged; -} - - -/* - * Mark specified partition as active. - */ -enum xpc_retval -xpc_mark_partition_active(struct xpc_partition *part) -{ - unsigned long irq_flags; - enum xpc_retval ret; - - - dev_dbg(xpc_part, "setting partition %d to ACTIVE\n", XPC_PARTID(part)); - - spin_lock_irqsave(&part->act_lock, irq_flags); - if (part->act_state == XPC_P_ACTIVATING) { - part->act_state = XPC_P_ACTIVE; - ret = xpcSuccess; - } else { - DBUG_ON(part->reason == xpcSuccess); - ret = part->reason; - } - spin_unlock_irqrestore(&part->act_lock, irq_flags); - - return ret; -} - - -/* - * Notify XPC that the partition is down. - */ -void -xpc_deactivate_partition(const int line, struct xpc_partition *part, - enum xpc_retval reason) -{ - unsigned long irq_flags; - - - spin_lock_irqsave(&part->act_lock, irq_flags); - - if (part->act_state == XPC_P_INACTIVE) { - XPC_SET_REASON(part, reason, line); - spin_unlock_irqrestore(&part->act_lock, irq_flags); - if (reason == xpcReactivating) { - /* we interrupt ourselves to reactivate partition */ - xpc_IPI_send_reactivate(part); - } - return; - } - if (part->act_state == XPC_P_DEACTIVATING) { - if ((part->reason == xpcUnloading && reason != xpcUnloading) || - reason == xpcReactivating) { - XPC_SET_REASON(part, reason, line); - } - spin_unlock_irqrestore(&part->act_lock, irq_flags); - return; - } - - part->act_state = XPC_P_DEACTIVATING; - XPC_SET_REASON(part, reason, line); - - spin_unlock_irqrestore(&part->act_lock, irq_flags); - - if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) { - xpc_request_partition_disengage(part); - xpc_IPI_send_disengage(part); - - /* set a timelimit on the disengage request */ - part->disengage_request_timeout = jiffies + - (xpc_disengage_request_timelimit * HZ); - part->disengage_request_timer.expires = - part->disengage_request_timeout; - add_timer(&part->disengage_request_timer); - } - - dev_dbg(xpc_part, "bringing partition %d down, reason = %d\n", - XPC_PARTID(part), reason); - - xpc_partition_going_down(part, reason); -} - - -/* - * Mark specified partition as inactive. - */ -void -xpc_mark_partition_inactive(struct xpc_partition *part) -{ - unsigned long irq_flags; - - - dev_dbg(xpc_part, "setting partition %d to INACTIVE\n", - XPC_PARTID(part)); - - spin_lock_irqsave(&part->act_lock, irq_flags); - part->act_state = XPC_P_INACTIVE; - spin_unlock_irqrestore(&part->act_lock, irq_flags); - part->remote_rp_pa = 0; -} - - -/* - * SAL has provided a partition and machine mask. The partition mask - * contains a bit for each even nasid in our partition. The machine - * mask contains a bit for each even nasid in the entire machine. - * - * Using those two bit arrays, we can determine which nasids are - * known in the machine. Each should also have a reserved page - * initialized if they are available for partitioning. - */ -void -xpc_discovery(void) -{ - void *remote_rp_base; - struct xpc_rsvd_page *remote_rp; - struct xpc_vars *remote_vars; - u64 remote_rp_pa; - u64 remote_vars_pa; - int region; - int region_size; - int max_regions; - int nasid; - struct xpc_rsvd_page *rp; - partid_t partid; - struct xpc_partition *part; - u64 *discovered_nasids; - enum xpc_retval ret; - - - remote_rp = xpc_kmalloc_cacheline_aligned(XPC_RP_HEADER_SIZE + - xp_nasid_mask_bytes, - GFP_KERNEL, &remote_rp_base); - if (remote_rp == NULL) { - return; - } - remote_vars = (struct xpc_vars *) remote_rp; - - - discovered_nasids = kzalloc(sizeof(u64) * xp_nasid_mask_words, - GFP_KERNEL); - if (discovered_nasids == NULL) { - kfree(remote_rp_base); - return; - } - - rp = (struct xpc_rsvd_page *) xpc_rsvd_page; - - /* - * The term 'region' in this context refers to the minimum number of - * nodes that can comprise an access protection grouping. The access - * protection is in regards to memory, IOI and IPI. - */ - max_regions = 64; - region_size = sn_region_size; - - switch (region_size) { - case 128: - max_regions *= 2; - case 64: - max_regions *= 2; - case 32: - max_regions *= 2; - region_size = 16; - DBUG_ON(!is_shub2()); - } - - for (region = 0; region < max_regions; region++) { - - if ((volatile int) xpc_exiting) { - break; - } - - dev_dbg(xpc_part, "searching region %d\n", region); - - for (nasid = (region * region_size * 2); - nasid < ((region + 1) * region_size * 2); - nasid += 2) { - - if ((volatile int) xpc_exiting) { - break; - } - - dev_dbg(xpc_part, "checking nasid %d\n", nasid); - - - if (XPC_NASID_IN_ARRAY(nasid, xpc_part_nasids)) { - dev_dbg(xpc_part, "PROM indicates Nasid %d is " - "part of the local partition; skipping " - "region\n", nasid); - break; - } - - if (!(XPC_NASID_IN_ARRAY(nasid, xpc_mach_nasids))) { - dev_dbg(xpc_part, "PROM indicates Nasid %d was " - "not on Numa-Link network at reset\n", - nasid); - continue; - } - - if (XPC_NASID_IN_ARRAY(nasid, discovered_nasids)) { - dev_dbg(xpc_part, "Nasid %d is part of a " - "partition which was previously " - "discovered\n", nasid); - continue; - } - - - /* pull over the reserved page structure */ - - ret = xpc_get_remote_rp(nasid, discovered_nasids, - remote_rp, &remote_rp_pa); - if (ret != xpcSuccess) { - dev_dbg(xpc_part, "unable to get reserved page " - "from nasid %d, reason=%d\n", nasid, - ret); - - if (ret == xpcLocalPartid) { - break; - } - continue; - } - - remote_vars_pa = remote_rp->vars_pa; - - partid = remote_rp->partid; - part = &xpc_partitions[partid]; - - - /* pull over the cross partition variables */ - - ret = xpc_get_remote_vars(remote_vars_pa, remote_vars); - if (ret != xpcSuccess) { - dev_dbg(xpc_part, "unable to get XPC variables " - "from nasid %d, reason=%d\n", nasid, - ret); - - XPC_DEACTIVATE_PARTITION(part, ret); - continue; - } - - if (part->act_state != XPC_P_INACTIVE) { - dev_dbg(xpc_part, "partition %d on nasid %d is " - "already activating\n", partid, nasid); - break; - } - - /* - * Register the remote partition's AMOs with SAL so it - * can handle and cleanup errors within that address - * range should the remote partition go down. We don't - * unregister this range because it is difficult to - * tell when outstanding writes to the remote partition - * are finished and thus when it is thus safe to - * unregister. This should not result in wasted space - * in the SAL xp_addr_region table because we should - * get the same page for remote_act_amos_pa after - * module reloads and system reboots. - */ - if (sn_register_xp_addr_region( - remote_vars->amos_page_pa, - PAGE_SIZE, 1) < 0) { - dev_dbg(xpc_part, "partition %d failed to " - "register xp_addr region 0x%016lx\n", - partid, remote_vars->amos_page_pa); - - XPC_SET_REASON(part, xpcPhysAddrRegFailed, - __LINE__); - break; - } - - /* - * The remote nasid is valid and available. - * Send an interrupt to that nasid to notify - * it that we are ready to begin activation. - */ - dev_dbg(xpc_part, "sending an interrupt to AMO 0x%lx, " - "nasid %d, phys_cpuid 0x%x\n", - remote_vars->amos_page_pa, - remote_vars->act_nasid, - remote_vars->act_phys_cpuid); - - if (XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars-> - version)) { - part->remote_amos_page_pa = - remote_vars->amos_page_pa; - xpc_mark_partition_disengaged(part); - xpc_cancel_partition_disengage_request(part); - } - xpc_IPI_send_activate(remote_vars); - } - } - - kfree(discovered_nasids); - kfree(remote_rp_base); -} - - -/* - * Given a partid, get the nasids owned by that partition from the - * remote partition's reserved page. - */ -enum xpc_retval -xpc_initiate_partid_to_nasids(partid_t partid, void *nasid_mask) -{ - struct xpc_partition *part; - u64 part_nasid_pa; - int bte_res; - - - part = &xpc_partitions[partid]; - if (part->remote_rp_pa == 0) { - return xpcPartitionDown; - } - - memset(nasid_mask, 0, XP_NASID_MASK_BYTES); - - part_nasid_pa = (u64) XPC_RP_PART_NASIDS(part->remote_rp_pa); - - bte_res = xp_bte_copy(part_nasid_pa, (u64) nasid_mask, - xp_nasid_mask_bytes, (BTE_NOTIFY | BTE_WACQUIRE), NULL); - - return xpc_map_bte_errors(bte_res); -} - diff --git a/arch/ia64/sn/kernel/xpnet.c b/arch/ia64/sn/kernel/xpnet.c deleted file mode 100644 index 007703c494a..00000000000 --- a/arch/ia64/sn/kernel/xpnet.c +++ /dev/null @@ -1,718 +0,0 @@ -/* - * This file is subject to the terms and conditions of the GNU General Public - * License. See the file "COPYING" in the main directory of this archive - * for more details. - * - * Copyright (C) 1999,2001-2005 Silicon Graphics, Inc. All rights reserved. - */ - - -/* - * Cross Partition Network Interface (XPNET) support - * - * XPNET provides a virtual network layered on top of the Cross - * Partition communication layer. - * - * XPNET provides direct point-to-point and broadcast-like support - * for an ethernet-like device. The ethernet broadcast medium is - * replaced with a point-to-point message structure which passes - * pointers to a DMA-capable block that a remote partition should - * retrieve and pass to the upper level networking layer. - * - */ - - -#include <linux/module.h> -#include <linux/kernel.h> -#include <linux/pci.h> -#include <linux/init.h> -#include <linux/ioport.h> -#include <linux/netdevice.h> -#include <linux/etherdevice.h> -#include <linux/delay.h> -#include <linux/ethtool.h> -#include <linux/mii.h> -#include <linux/smp.h> -#include <linux/string.h> -#include <asm/sn/bte.h> -#include <asm/sn/io.h> -#include <asm/sn/sn_sal.h> -#include <asm/types.h> -#include <asm/atomic.h> -#include <asm/sn/xp.h> - - -/* - * The message payload transferred by XPC. - * - * buf_pa is the physical address where the DMA should pull from. - * - * NOTE: for performance reasons, buf_pa should _ALWAYS_ begin on a - * cacheline boundary. To accomplish this, we record the number of - * bytes from the beginning of the first cacheline to the first useful - * byte of the skb (leadin_ignore) and the number of bytes from the - * last useful byte of the skb to the end of the last cacheline - * (tailout_ignore). - * - * size is the number of bytes to transfer which includes the skb->len - * (useful bytes of the senders skb) plus the leadin and tailout - */ -struct xpnet_message { - u16 version; /* Version for this message */ - u16 embedded_bytes; /* #of bytes embedded in XPC message */ - u32 magic; /* Special number indicating this is xpnet */ - u64 buf_pa; /* phys address of buffer to retrieve */ - u32 size; /* #of bytes in buffer */ - u8 leadin_ignore; /* #of bytes to ignore at the beginning */ - u8 tailout_ignore; /* #of bytes to ignore at the end */ - unsigned char data; /* body of small packets */ -}; - -/* - * Determine the size of our message, the cacheline aligned size, - * and then the number of message will request from XPC. - * - * XPC expects each message to exist in an individual cacheline. - */ -#define XPNET_MSG_SIZE (L1_CACHE_BYTES - XPC_MSG_PAYLOAD_OFFSET) -#define XPNET_MSG_DATA_MAX \ - (XPNET_MSG_SIZE - (u64)(&((struct xpnet_message *)0)->data)) -#define XPNET_MSG_ALIGNED_SIZE (L1_CACHE_ALIGN(XPNET_MSG_SIZE)) -#define XPNET_MSG_NENTRIES (PAGE_SIZE / XPNET_MSG_ALIGNED_SIZE) - - -#define XPNET_MAX_KTHREADS (XPNET_MSG_NENTRIES + 1) -#define XPNET_MAX_IDLE_KTHREADS (XPNET_MSG_NENTRIES + 1) - -/* - * Version number of XPNET implementation. XPNET can always talk to versions - * with same major #, and never talk to versions with a different version. - */ -#define _XPNET_VERSION(_major, _minor) (((_major) << 4) | (_minor)) -#define XPNET_VERSION_MAJOR(_v) ((_v) >> 4) -#define XPNET_VERSION_MINOR(_v) ((_v) & 0xf) - -#define XPNET_VERSION _XPNET_VERSION(1,0) /* version 1.0 */ -#define XPNET_VERSION_EMBED _XPNET_VERSION(1,1) /* version 1.1 */ -#define XPNET_MAGIC 0x88786984 /* "XNET" */ - -#define XPNET_VALID_MSG(_m) \ - ((XPNET_VERSION_MAJOR(_m->version) == XPNET_VERSION_MAJOR(XPNET_VERSION)) \ - && (msg->magic == XPNET_MAGIC)) - -#define XPNET_DEVICE_NAME "xp0" - - -/* - * When messages are queued with xpc_send_notify, a kmalloc'd buffer - * of the following type is passed as a notification cookie. When the - * notification function is called, we use the cookie to decide - * whether all outstanding message sends have completed. The skb can - * then be released. - */ -struct xpnet_pending_msg { - struct list_head free_list; - struct sk_buff *skb; - atomic_t use_count; -}; - -/* driver specific structure pointed to by the device structure */ -struct xpnet_dev_private { - struct net_device_stats stats; -}; - -struct net_device *xpnet_device; - -/* - * When we are notified of other partitions activating, we add them to - * our bitmask of partitions to which we broadcast. - */ -static u64 xpnet_broadcast_partitions; -/* protect above */ -static DEFINE_SPINLOCK(xpnet_broadcast_lock); - -/* - * Since the Block Transfer Engine (BTE) is being used for the transfer - * and it relies upon cache-line size transfers, we need to reserve at - * least one cache-line for head and tail alignment. The BTE is - * limited to 8MB transfers. - * - * Testing has shown that changing MTU to greater than 64KB has no effect - * on TCP as the two sides negotiate a Max Segment Size that is limited - * to 64K. Other protocols May use packets greater than this, but for - * now, the default is 64KB. - */ -#define XPNET_MAX_MTU (0x800000UL - L1_CACHE_BYTES) -/* 32KB has been determined to be the ideal */ -#define XPNET_DEF_MTU (0x8000UL) - - -/* - * The partition id is encapsulated in the MAC address. The following - * define locates the octet the partid is in. - */ -#define XPNET_PARTID_OCTET 1 -#define XPNET_LICENSE_OCTET 2 - - -/* - * Define the XPNET debug device structure that is to be used with dev_dbg(), - * dev_err(), dev_warn(), and dev_info(). - */ -struct device_driver xpnet_dbg_name = { - .name = "xpnet" -}; - -struct device xpnet_dbg_subname = { - .bus_id = {0}, /* set to "" */ - .driver = &xpnet_dbg_name -}; - -struct device *xpnet = &xpnet_dbg_subname; - -/* - * Packet was recevied by XPC and forwarded to us. - */ -static void -xpnet_receive(partid_t partid, int channel, struct xpnet_message *msg) -{ - struct sk_buff *skb; - bte_result_t bret; - struct xpnet_dev_private *priv = - (struct xpnet_dev_private *) xpnet_device->priv; - - - if (!XPNET_VALID_MSG(msg)) { - /* - * Packet with a different XPC version. Ignore. - */ - xpc_received(partid, channel, (void *) msg); - - priv->stats.rx_errors++; - - return; - } - dev_dbg(xpnet, "received 0x%lx, %d, %d, %d\n", msg->buf_pa, msg->size, - msg->leadin_ignore, msg->tailout_ignore); - - - /* reserve an extra cache line */ - skb = dev_alloc_skb(msg->size + L1_CACHE_BYTES); - if (!skb) { - dev_err(xpnet, "failed on dev_alloc_skb(%d)\n", - msg->size + L1_CACHE_BYTES); - - xpc_received(partid, channel, (void *) msg); - - priv->stats.rx_errors++; - - return; - } - - /* - * The allocated skb has some reserved space. - * In order to use bte_copy, we need to get the - * skb->data pointer moved forward. - */ - skb_reserve(skb, (L1_CACHE_BYTES - ((u64)skb->data & - (L1_CACHE_BYTES - 1)) + - msg->leadin_ignore)); - - /* - * Update the tail pointer to indicate data actually - * transferred. - */ - skb_put(skb, (msg->size - msg->leadin_ignore - msg->tailout_ignore)); - - /* - * Move the data over from the the other side. - */ - if ((XPNET_VERSION_MINOR(msg->version) == 1) && - (msg->embedded_bytes != 0)) { - dev_dbg(xpnet, "copying embedded message. memcpy(0x%p, 0x%p, " - "%lu)\n", skb->data, &msg->data, - (size_t) msg->embedded_bytes); - - memcpy(skb->data, &msg->data, (size_t) msg->embedded_bytes); - } else { - dev_dbg(xpnet, "transferring buffer to the skb->data area;\n\t" - "bte_copy(0x%p, 0x%p, %hu)\n", (void *)msg->buf_pa, - (void *)__pa((u64)skb->data & ~(L1_CACHE_BYTES - 1)), - msg->size); - - bret = bte_copy(msg->buf_pa, - __pa((u64)skb->data & ~(L1_CACHE_BYTES - 1)), - msg->size, (BTE_NOTIFY | BTE_WACQUIRE), NULL); - - if (bret != BTE_SUCCESS) { - // >>> Need better way of cleaning skb. Currently skb - // >>> appears in_use and we can't just call - // >>> dev_kfree_skb. - dev_err(xpnet, "bte_copy(0x%p, 0x%p, 0x%hx) returned " - "error=0x%x\n", (void *)msg->buf_pa, - (void *)__pa((u64)skb->data & - ~(L1_CACHE_BYTES - 1)), - msg->size, bret); - - xpc_received(partid, channel, (void *) msg); - - priv->stats.rx_errors++; - - return; - } - } - - dev_dbg(xpnet, "<skb->head=0x%p skb->data=0x%p skb->tail=0x%p " - "skb->end=0x%p skb->len=%d\n", (void *) skb->head, - (void *) skb->data, (void *) skb->tail, (void *) skb->end, - skb->len); - - skb->dev = xpnet_device; - skb->protocol = eth_type_trans(skb, xpnet_device); - skb->ip_summed = CHECKSUM_UNNECESSARY; - - dev_dbg(xpnet, "passing skb to network layer; \n\tskb->head=0x%p " - "skb->data=0x%p skb->tail=0x%p skb->end=0x%p skb->len=%d\n", - (void *) skb->head, (void *) skb->data, (void *) skb->tail, - (void *) skb->end, skb->len); - - - xpnet_device->last_rx = jiffies; - priv->stats.rx_packets++; - priv->stats.rx_bytes += skb->len + ETH_HLEN; - - netif_rx_ni(skb); - xpc_received(partid, channel, (void *) msg); -} - - -/* - * This is the handler which XPC calls during any sort of change in - * state or message reception on a connection. - */ -static void -xpnet_connection_activity(enum xpc_retval reason, partid_t partid, int channel, - void *data, void *key) -{ - long bp; - - - DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS); - DBUG_ON(channel != XPC_NET_CHANNEL); - - switch(reason) { - case xpcMsgReceived: /* message received */ - DBUG_ON(data == NULL); - - xpnet_receive(partid, channel, (struct xpnet_message *) data); - break; - - case xpcConnected: /* connection completed to a partition */ - spin_lock_bh(&xpnet_broadcast_lock); - xpnet_broadcast_partitions |= 1UL << (partid -1 ); - bp = xpnet_broadcast_partitions; - spin_unlock_bh(&xpnet_broadcast_lock); - - netif_carrier_on(xpnet_device); - - dev_dbg(xpnet, "%s connection created to partition %d; " - "xpnet_broadcast_partitions=0x%lx\n", - xpnet_device->name, partid, bp); - break; - - default: - spin_lock_bh(&xpnet_broadcast_lock); - xpnet_broadcast_partitions &= ~(1UL << (partid -1 )); - bp = xpnet_broadcast_partitions; - spin_unlock_bh(&xpnet_broadcast_lock); - - if (bp == 0) { - netif_carrier_off(xpnet_device); - } - - dev_dbg(xpnet, "%s disconnected from partition %d; " - "xpnet_broadcast_partitions=0x%lx\n", - xpnet_device->name, partid, bp); - break; - - } -} - - -static int -xpnet_dev_open(struct net_device *dev) -{ - enum xpc_retval ret; - - - dev_dbg(xpnet, "calling xpc_connect(%d, 0x%p, NULL, %ld, %ld, %d, " - "%d)\n", XPC_NET_CHANNEL, xpnet_connection_activity, - XPNET_MSG_SIZE, XPNET_MSG_NENTRIES, XPNET_MAX_KTHREADS, - XPNET_MAX_IDLE_KTHREADS); - - ret = xpc_connect(XPC_NET_CHANNEL, xpnet_connection_activity, NULL, - XPNET_MSG_SIZE, XPNET_MSG_NENTRIES, - XPNET_MAX_KTHREADS, XPNET_MAX_IDLE_KTHREADS); - if (ret != xpcSuccess) { - dev_err(xpnet, "ifconfig up of %s failed on XPC connect, " - "ret=%d\n", dev->name, ret); - - return -ENOMEM; - } - - dev_dbg(xpnet, "ifconfig up of %s; XPC connected\n", dev->name); - - return 0; -} - - -static int -xpnet_dev_stop(struct net_device *dev) -{ - xpc_disconnect(XPC_NET_CHANNEL); - - dev_dbg(xpnet, "ifconfig down of %s; XPC disconnected\n", dev->name); - - return 0; -} - - -static int -xpnet_dev_change_mtu(struct net_device *dev, int new_mtu) -{ - /* 68 comes from min TCP+IP+MAC header */ - if ((new_mtu < 68) || (new_mtu > XPNET_MAX_MTU)) { - dev_err(xpnet, "ifconfig %s mtu %d failed; value must be " - "between 68 and %ld\n", dev->name, new_mtu, - XPNET_MAX_MTU); - return -EINVAL; - } - - dev->mtu = new_mtu; - dev_dbg(xpnet, "ifconfig %s mtu set to %d\n", dev->name, new_mtu); - return 0; -} - - -/* - * Required for the net_device structure. - */ -static int -xpnet_dev_set_config(struct net_device *dev, struct ifmap *new_map) -{ - return 0; -} - - -/* - * Return statistics to the caller. - */ -static struct net_device_stats * -xpnet_dev_get_stats(struct net_device *dev) -{ - struct xpnet_dev_private *priv; - - - priv = (struct xpnet_dev_private *) dev->priv; - - return &priv->stats; -} - - -/* - * Notification that the other end has received the message and - * DMA'd the skb information. At this point, they are done with - * our side. When all recipients are done processing, we - * release the skb and then release our pending message structure. - */ -static void -xpnet_send_completed(enum xpc_retval reason, partid_t partid, int channel, - void *__qm) -{ - struct xpnet_pending_msg *queued_msg = - (struct xpnet_pending_msg *) __qm; - - - DBUG_ON(queued_msg == NULL); - - dev_dbg(xpnet, "message to %d notified with reason %d\n", - partid, reason); - - if (atomic_dec_return(&queued_msg->use_count) == 0) { - dev_dbg(xpnet, "all acks for skb->head=-x%p\n", - (void *) queued_msg->skb->head); - - dev_kfree_skb_any(queued_msg->skb); - kfree(queued_msg); - } -} - - -/* - * Network layer has formatted a packet (skb) and is ready to place it - * "on the wire". Prepare and send an xpnet_message to all partitions - * which have connected with us and are targets of this packet. - * - * MAC-NOTE: For the XPNET driver, the MAC address contains the - * destination partition_id. If the destination partition id word - * is 0xff, this packet is to broadcast to all partitions. - */ -static int -xpnet_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev) -{ - struct xpnet_pending_msg *queued_msg; - enum xpc_retval ret; - struct xpnet_message *msg; - u64 start_addr, end_addr; - long dp; - u8 second_mac_octet; - partid_t dest_partid; - struct xpnet_dev_private *priv; - u16 embedded_bytes; - - - priv = (struct xpnet_dev_private *) dev->priv; - - - dev_dbg(xpnet, ">skb->head=0x%p skb->data=0x%p skb->tail=0x%p " - "skb->end=0x%p skb->len=%d\n", (void *) skb->head, - (void *) skb->data, (void *) skb->tail, (void *) skb->end, - skb->len); - - - /* - * The xpnet_pending_msg tracks how many outstanding - * xpc_send_notifies are relying on this skb. When none - * remain, release the skb. - */ - queued_msg = kmalloc(sizeof(struct xpnet_pending_msg), GFP_ATOMIC); - if (queued_msg == NULL) { - dev_warn(xpnet, "failed to kmalloc %ld bytes; dropping " - "packet\n", sizeof(struct xpnet_pending_msg)); - - priv->stats.tx_errors++; - - return -ENOMEM; - } - - - /* get the beginning of the first cacheline and end of last */ - start_addr = ((u64) skb->data & ~(L1_CACHE_BYTES - 1)); - end_addr = L1_CACHE_ALIGN((u64) skb->tail); - - /* calculate how many bytes to embed in the XPC message */ - embedded_bytes = 0; - if (unlikely(skb->len <= XPNET_MSG_DATA_MAX)) { - /* skb->data does fit so embed */ - embedded_bytes = skb->len; - } - - - /* - * Since the send occurs asynchronously, we set the count to one - * and begin sending. Any sends that happen to complete before - * we are done sending will not free the skb. We will be left - * with that task during exit. This also handles the case of - * a packet destined for a partition which is no longer up. - */ - atomic_set(&queued_msg->use_count, 1); - queued_msg->skb = skb; - - - second_mac_octet = skb->data[XPNET_PARTID_OCTET]; - if (second_mac_octet == 0xff) { - /* we are being asked to broadcast to all partitions */ - dp = xpnet_broadcast_partitions; - } else if (second_mac_octet != 0) { - dp = xpnet_broadcast_partitions & - (1UL << (second_mac_octet - 1)); - } else { - /* 0 is an invalid partid. Ignore */ - dp = 0; - } - dev_dbg(xpnet, "destination Partitions mask (dp) = 0x%lx\n", dp); - - /* - * If we wanted to allow promiscous mode to work like an - * unswitched network, this would be a good point to OR in a - * mask of partitions which should be receiving all packets. - */ - - /* - * Main send loop. - */ - for (dest_partid = 1; dp && dest_partid < XP_MAX_PARTITIONS; - dest_partid++) { - - - if (!(dp & (1UL << (dest_partid - 1)))) { - /* not destined for this partition */ - continue; - } - - /* remove this partition from the destinations mask */ - dp &= ~(1UL << (dest_partid - 1)); - - - /* found a partition to send to */ - - ret = xpc_allocate(dest_partid, XPC_NET_CHANNEL, - XPC_NOWAIT, (void **)&msg); - if (unlikely(ret != xpcSuccess)) { - continue; - } - - msg->embedded_bytes = embedded_bytes; - if (unlikely(embedded_bytes != 0)) { - msg->version = XPNET_VERSION_EMBED; - dev_dbg(xpnet, "calling memcpy(0x%p, 0x%p, 0x%lx)\n", - &msg->data, skb->data, (size_t) embedded_bytes); - memcpy(&msg->data, skb->data, (size_t) embedded_bytes); - } else { - msg->version = XPNET_VERSION; - } - msg->magic = XPNET_MAGIC; - msg->size = end_addr - start_addr; - msg->leadin_ignore = (u64) skb->data - start_addr; - msg->tailout_ignore = end_addr - (u64) skb->tail; - msg->buf_pa = __pa(start_addr); - - dev_dbg(xpnet, "sending XPC message to %d:%d\nmsg->buf_pa=" - "0x%lx, msg->size=%u, msg->leadin_ignore=%u, " - "msg->tailout_ignore=%u\n", dest_partid, - XPC_NET_CHANNEL, msg->buf_pa, msg->size, - msg->leadin_ignore, msg->tailout_ignore); - - - atomic_inc(&queued_msg->use_count); - - ret = xpc_send_notify(dest_partid, XPC_NET_CHANNEL, msg, - xpnet_send_completed, queued_msg); - if (unlikely(ret != xpcSuccess)) { - atomic_dec(&queued_msg->use_count); - continue; - } - - } - - if (atomic_dec_return(&queued_msg->use_count) == 0) { - dev_dbg(xpnet, "no partitions to receive packet destined for " - "%d\n", dest_partid); - - - dev_kfree_skb(skb); - kfree(queued_msg); - } - - priv->stats.tx_packets++; - priv->stats.tx_bytes += skb->len; - - return 0; -} - - -/* - * Deal with transmit timeouts coming from the network layer. - */ -static void -xpnet_dev_tx_timeout (struct net_device *dev) -{ - struct xpnet_dev_private *priv; - - - priv = (struct xpnet_dev_private *) dev->priv; - - priv->stats.tx_errors++; - return; -} - - -static int __init -xpnet_init(void) -{ - int i; - u32 license_num; - int result = -ENOMEM; - - - if (!ia64_platform_is("sn2")) { - return -ENODEV; - } - - dev_info(xpnet, "registering network device %s\n", XPNET_DEVICE_NAME); - - /* - * use ether_setup() to init the majority of our device - * structure and then override the necessary pieces. - */ - xpnet_device = alloc_netdev(sizeof(struct xpnet_dev_private), - XPNET_DEVICE_NAME, ether_setup); - if (xpnet_device == NULL) { - return -ENOMEM; - } - - netif_carrier_off(xpnet_device); - - xpnet_device->mtu = XPNET_DEF_MTU; - xpnet_device->change_mtu = xpnet_dev_change_mtu; - xpnet_device->open = xpnet_dev_open; - xpnet_device->get_stats = xpnet_dev_get_stats; - xpnet_device->stop = xpnet_dev_stop; - xpnet_device->hard_start_xmit = xpnet_dev_hard_start_xmit; - xpnet_device->tx_timeout = xpnet_dev_tx_timeout; - xpnet_device->set_config = xpnet_dev_set_config; - - /* - * Multicast assumes the LSB of the first octet is set for multicast - * MAC addresses. We chose the first octet of the MAC to be unlikely - * to collide with any vendor's officially issued MAC. - */ - xpnet_device->dev_addr[0] = 0xfe; - xpnet_device->dev_addr[XPNET_PARTID_OCTET] = sn_partition_id; - license_num = sn_partition_serial_number_val(); - for (i = 3; i >= 0; i--) { - xpnet_device->dev_addr[XPNET_LICENSE_OCTET + i] = - license_num & 0xff; - license_num = license_num >> 8; - } - - /* - * ether_setup() sets this to a multicast device. We are - * really not supporting multicast at this time. - */ - xpnet_device->flags &= ~IFF_MULTICAST; - - /* - * No need to checksum as it is a DMA transfer. The BTE will - * report an error if the data is not retrievable and the - * packet will be dropped. - */ - xpnet_device->features = NETIF_F_NO_CSUM; - - result = register_netdev(xpnet_device); - if (result != 0) { - free_netdev(xpnet_device); - } - - return result; -} -module_init(xpnet_init); - - -static void __exit -xpnet_exit(void) -{ - dev_info(xpnet, "unregistering network device %s\n", - xpnet_device[0].name); - - unregister_netdev(xpnet_device); - - free_netdev(xpnet_device); -} -module_exit(xpnet_exit); - - -MODULE_AUTHOR("Silicon Graphics, Inc."); -MODULE_DESCRIPTION("Cross Partition Network adapter (XPNET)"); -MODULE_LICENSE("GPL"); - |